The number and rate of euploid blastocysts in women undergoing IVF/ICSI cycles are strongly dependent on ovarian reserve and female age

Abstract

Study question: Can the possibility of having at least one euploid blastocyst for embryo transfer and the total number of euploid blastocysts be predicted for couples before they enter the IVF programme?

Summary answer: Ovarian reserve and female age are the most important predictors of having at least one euploid blastocyst and the total number of euploid blastocysts.

What is known already: The blastocyst euploidy rate among women undergoing ART has already been shown to significantly decrease with increasing female age, and the total number of euploid embryos is dependent on the blastocyst cohort size. However, the vast majority of published studies are based on retrospective analysis of data.

Study design, size, duration: This prospective analysis included 847 consecutively enrolled couples approaching their first preimplantation genetic testing for aneuploidies (PGT-A) cycle between 2017 and 2020. Only couples for whom ejaculated sperm was available and women with a BMI of <35 kg/m2 were included in the study. Only the first cycle was included for each patient.

Participants/materials, setting, methods: The study was conducted at an IVF centre where, for all patients, the planned treatment was to obtain embryos at the blastocyst stage for the PGT-A programme. The impact of the following covariates was investigated: a woman's serum AMH level, age, height, weight and BMI and a man's age, height, weight, BMI, sperm volume and sperm motility and morphology. The analysis was performed with a machine learning (ML) approach. Models were fit on the training set (677 patients) and their predictive performance was then evaluated on the test set (170 patients).

Main results and role of chance: After ovarian stimulation and oocyte insemination, 40.1% of couples had at least one blastocyst available for the PGT-A. Of 1068 blastocysts analysed, 33.6% were euploid. Two distinct ML models were fit: one for the probability of having at least one euploid blastocyst and one for the number of euploid blastocysts obtained. In the training set of patients, the variable importance plots of both models indicated that AMH and the woman's age are by far the most important predictors. Specifically, a positive association between the outcome and AMH and a negative association between the outcome and female age appeared. Gradient-boosted modelling offers a greater predictive performance than generalized additive models (GAMs).

Limitations, reasons for caution: The study was performed based on data from a single centre. While this provides a robust set of data with a constant ART process and laboratory practice, the model might be suitable only for the evaluated population, which may limit the generalization of the model to other populations.

Wider implications of the findings: ML models indicate that for couples entering the IVF/PGT-A programme, ovarian reserve, which is known to vary with age, is the most important predictor of having at least one euploid embryo. According to the GAM, the probability of a 30-year-old woman having at least one euploid embryo is 28% or 47% if her AMH level is 1 or 3 ng/ml, respectively; if the woman is 40 years old, this probability is 18% with an AMH of 1 ng/ml and 30% with an AMH of 3 ng/ml.

Study funding/competing interest(s): This study was supported by an unrestricted grant from Gedeon Richter. The authors declared no conflict of interests.

Trial registration number: N/A.

Keywords: AMH; anti-Müllerian hormone; blastocyst euploidy; female age; machine learning; ovarian reserve; predictive models.