Improved clinical outcomes after non-invasive oocyte selection and Day 3 eSET in ICSI patients

The primary objective was to investigate whether, in a larger patient cohort of patients with a Day 3 eSET, CPR could be increased by analyzing the gene expression profile of 3 predictive genes in the cumulus cells compared to patients with morphology-based embryo selection only. Secondary endpoints were 1. LBR, 2. CPR and LBR in subgroups according to age and number of transferable embryos, and 3. time-to-pregnancy.

Women up to 39 years old, stimulated with HP-hMG (Menopur®, Ferring Pharmaceuticals, St. Prex, Switzerland), in a GnRH antagonist protocol, who presented at a tertiary referral hospital between October 2013 and April 2019, scheduled for ICSI and fresh single embryo transfer on Day 3, were eligible for this interventional, non-randomized, assessor-blinded cohort study with one experimental arm and one control arm. Patients with polycystic ovary syndrome (PCOS, Rotterdam 2003 criteria [28]) and/or severe male infertility were excluded from the study. There was no statistical difference in infertility indication (male, female or mixed infertility) between the arms of the study. Patients enrolled in other studies or scheduled for PGT (pre-implantation genetic testing) were also excluded. Patients with only one or no transferable embryo on Day 3 were considered drop-outs.

The design of the predictive model and the selection of the three genes was described before in a manuscript reporting the results of our pilot study [1]. In brief, more than 140 Affymetrix arrays on individual cumulus cells (CC) of SET ICSI patients were used for transcriptome analysis with embryo development and live birth as endpoints. The different microarray analyses revealed a multitude of potential embryo development and pregnancy predicting genes (unpublished data). From these lists, 23 predictive genes were validated over time in independent biological replicates using qRT-PCR, with a focus on genes predicting pregnancy. Several of these qPCR studies were published [2, 36‐38]. While the expression of many genes could be related to oocyte competence, the two main challenges were: finding the strongest combination of genes and finding a model predictive for live birth. This validation strategy together with intrapatient comparisons have led to the current pregnancy prediction model (AUC 0,8081; accuracy 80%). EFNB2, SASH1, and CAMK1D have been linked to cell expansion [9], the Toll-like receptor 4 pathway [11], and the calcium pathway, respectively. In our study, EFNB2 and SASH1 expression were positive correlated and CAMK1D exon 1 expression was negatively correlated with clinical pregnancy ([1], see Supplementary Figure 1).

This mathematical model, comprising only gene expression results for predicting clinical pregnancy and live birth, was previously named Corona Test [1] and in this manuscript is named cumulus cell test. This test will be offered under the name “Aurora Test” (The name was changed because of the current Corona virus pandemic).

Cumulus cells (approx. 1000–30,000 cells/oocyte, extrapolated from total RNA measured with BioAnalyzer Pico Chip, Agilent) were collected after single-oocyte denudation using Cumulase (Origio, CooperSurgical) [34]. The individual oocyte denudation procedure (on average 8 oocytes per cycle) in the experimental arm required 15–30 min extra handling time compared to grouped oocyte denudation, depending on the number of oocytes per cycle. Within 4 h after oocyte retrieval, the cumulus cells were removed and ICSI was performed immediately thereafter. Cumulus samples of all fertilized oocytes were analyzed prospectively at the Follicle Biology laboratory of Vrije Universiteit Brussel - Universitair Ziekenhuis Brussel, Belgium, for all 113 patients in the experimental arm on Day 1 or 2 after oocyte retrieval. The lab performing the cumulus cell test was blinded for the morphological quality scoring of the embryos. Total RNA extraction on cumulus cells was performed with the RNeasy Micro kit (Qiagen, The Netherlands) on the Qiacube (Qiagen) and reversed transcription was performed with the iScript cDNA synthesis kit (BioRad, Belgium). cDNA was frozen at − 80 °C until further qPCR analysis. The three specific genes (EFNB2, SASH1, and CAMK1D) were analyzed together with two endogenous control genes, UBC and B2M. The mean of B2M and UBC expression was used as the normalization factor. All PCR quantifications (LC480, Roche Diagnostics) were performed in triplicate (for the specific genes EFNB2, CAMK1D and SASH1) or duplicate (for the endogenous controls B2M and UBC) for the samples, and in triplicate for the calibrators and negative controls. The average coefficient of variation was < 0,1 Cp for all assays applied. The mean laboratory turnaround time from the start of the sample processing by nucleic acid extraction and qRT-PCR up to the completion of the final report of the analysis was on average 8 h.

The normalized expression levels of the three genes were used to calculate a cumulus cell test ranking. Ranking data were reported to the embryology lab in the morning of Day 3. Embryos underwent the standard morphological evaluation comprising the scoring of fertilization, Day 2 embryo quality and full embryo grading on Day 3, as described previously [31]. Among the embryos that were morphologically eligible for transfer, the embryo with the highest cumulus cell test rank was selected by the embryologist for an eSET on Day 3.

The primary outcome measure was clinical pregnancy defined as the ultrasonographic visualization of a fetal sac at week 7 or later with normal fetal heartbeat. It also includes ectopic pregnancy [40, 41].

The secondary outcome measure was live birth defined as the complete expulsion or extraction from a woman of a product of fertilization, after 22 completed weeks of gestational age; which, after such separation, breathes or shows any other evidence of life, such as heart beat, umbilical cord pulsation or definite movement of voluntary muscles, irrespective of whether the umbilical cord has been cut or the placenta is attached [41].

Another secondary outcome measure was time-to-pregnancy defined as the time taken to establish a pregnancy, measured in months or in numbers of menstrual cycles [41]. In case of artificial ART cycles, it was measured in number of embryo transfer cycles.

An absolute increase of CPR of 25% points in the experimental arm over the control arm was assumed using a weighted generalized linear model for CPR and LBR, in which the weights given to the control patients were proportional to the number of patients in the experimental arm with the cumulus cell test in each corresponding group based on transferable embryos and age category. Sample size calculation showed that, if we considered a 95% power, and 20% variability in a two-sided test, because the morphological evaluation is performed by several embryologists, 107 informative patients were needed in the experimental arm, after all drop-outs, to show a 25% increase of CPR after fresh eSET. To prevent eventual confounding, all statistical analyses were stratified by age of the woman, and number of transferable embryos available on Day 3.

Patient characteristics were compared using the Wilcoxon rank sum test.

Information on cumulative CPR and LBR was obtained for all frozen embryo transfers. When considering the fresh and frozen transfers within 1 year from the start of treatment, the time-to-pregnancy was compared between the two arms using Kaplan-Meier curves. For this analysis, 113 exactly matched controls (matched for age, number of transferable embryos and closest in time to the treated case), out of the available 520 controls needed to be used.

This was an interventional, non-randomized, assessor-blinded cohort study with 633 patients, 113 patients in the experimental arm with the cumulus cell test and 520 patients in the control arm (Fig. 1). The study was designed to validate a 25% increase in CPR by ranking transferable embryos based on gene expression in cumulus cells. Secondary endpoint was to achieve a significant increase in LBR in the experimental arm. The number of controls was 4.7 times higher than the number of cases to ensure reliable data for CPR and LBR in the control arm.

The strict inclusion and exclusion criteria considerably limited the number of patients eligible for this study. The most important factors limiting patient inclusion were: (i) one specific stimulation protocol (GnRH antagonist with HP-hMG), and (ii) one specific embryo transfer regime (fresh eSET Day 3). Patients enrolled in other clinical studies performed at Universitair Ziekenhuis Brussel were excluded from this study.

The clinical patient characteristics in the experimental arm and the control arm were similar for most parameters except for the total stimulation dose and the number of transferable embryos on Day 3 (Table 1). Patients in the control arm received 2048 IU of HP-hMG vs 1900 IU in the experimental arm (p = 0.04). The number of transferable embryos was higher in the experimental arm (4.4 vs 3.6 in the control group, p = 0.0008). While there is a difference in the number of transferable embryos, the amount of top quality (EQ. 1) and high quality (EQ. 2) embryos is similar in all three patient groups (Table 1 and Supplementary Table 1). Histogram plots of the main patient characteristics (Figs. 2 and 3) show the three-dimensional and two-dimensional distributions and confirm the similarity between the two arms.

The CPR was 69/113 (61.1%) in the experimental arm and 153/520 (29.4%, weighted average) in the control arm (p < 0.0001) (Fig. 4). Patients are subcategorized for age and number of transferable embryos (3 age and 3 embryo quality categories: a total of 9 categories). With the weighted average the proportion of each group is taken into account to calculate the weighted averages for the clinical outcomes. As such, the contribution of each of the 9 subcategories is equal in the test group and the control group.

The LBR in the experimental arm was 56/113 (49.6%) vs 135/509 (26.7%, weighted average) in the control arm (p < 0.0001) (Fig. 4). As far as LBR is concerned, no patients were lost to follow-up in the experimental arm, whereas eleven patients were lost to live birth follow-up in the control arm. The majority of the 56 deliveries in the experimental arm were singleton live-born neonates. Two pregnancies resulted in a monozygotic twin birth. There were no stillbirths.

The experimental arm was further analyzed with respect to the number of transferable embryos in three subgroups (2, 3–4 and > 4 transferable embryos) and with respect to age in two subgroups (younger than 35 years, and 35–39 years old).

In the two subgroups with 2 and > 4 embryos, CPR was 66 and 67% and LBR was 53 and 54%, respectively. In the subgroup with 3–4 transferable embryos CPR and LBR dropped to 52% (p = 0,3) and 43% (p = 0,5), respectively. However, the CPR and LBR were not significantly lower (left side Fig. 5).

CPR was almost identical in both age subgroups (60% versus 62%; p = 0,9), while LBR was higher in the younger patient cohort versus the older patient cohort (55% versus 42%: p = 0,1), although this difference was not statistically significant.

Time-to-pregnancy was evaluated by Kaplan-Meier analysis for the 113 patients in the experimental arm vs 113 exact matched controls for all patients with > 2, or > 3, or > 4, or > 5, or > 6 transferable embryos. Time-to-pregnancy was significantly shorter in the experimental arm for all patients (p < 0.0001, Fig. 6). As an example, when considering all patients with at least two transferable embryos, three additional Day 3 transfers were needed in the control arm to achieve a clinical pregnancy compared with the patients in the experimental arm (Fig. 6).