L26/P-626 Factors affecting clinical pregnancy loss following euploid embryo transfer in PGT-A cycles: A machine learning approach
Human Reproduction · 01.07.2026 · 41 (Supplement_1)
DOI: 10.1093/humrep/deag083.956 ↗Kurz gefasst
Fachartikel in der Zeitschrift Human Reproduction, online erschienen am 08.07.2026. Verfasst von J.A. Ortiz und 6 weiteren Autor:innen.
Abstract (Originalsprache)
Study question What factors are associated with clinical pregnancy loss after euploid embryo transfer in PGT-A cycles? Summary answer Maternal-related factors, including age, BMI, AMH and progesterone levels, and uterine abnormalities, are the main predictors of clinical pregnancy loss after PGT-A cycles. What is known already Clinical pregnancy loss, defined as miscarriage after ultrasound confirmation of a gestational sac, affects approximately 10–25% of IVF pregnancies. Although chromosomal abnormalities are a major cause, clinical pregnancy loss still occurs after euploid embryo transfer following PGT-A, suggesting the involvement of non-chromosomal factors. Previous studies based on conventional statistical approaches have yielded inconsistent or inconclusive results. Consequently, alternative analytical strategies combining classical statistics with artificial intelligence may improve identification of factors associated with clinical pregnancy loss after PGT-A. Study design, size, duration This retrospective observational study included 1,216 embryo transfers from 911 PGT-A cycles performed between January 2017 and December 2024. Trophectoderm biopsies were conducted on D5, D6 or D7 blastocysts and analysed by next-generation sequencing using Illumina platforms (VeriSeq Illumina®, San Diego, CA, USA or EmbryoMap®, Vitrolife, Gothenburg, Sweden). All biopsied embryos were vitrified and transferred in subsequent frozen embryo transfer cycles. Participants/materials, setting, methods Indications for PGT-A included advanced maternal age, abnormal sperm FISH, chromosomal abnormalities in previous offspring, repeated miscarriages, and recurrent implantation failure. Clinical outcomes and 38 potential predictors related to patients or donors, embryos, biopsy procedures, ovarian stimulation, endometrial factors, and adjuvant treatments were recorded. Classical statistical analyses and machine learning algorithms were performed using R software (version 4.4.2). Main results and the role of chance Mean age was 39.8 ± 3.8 years in patients, 35.1 ± 6.3 years in patients/oocyte donors undergoing ovarian stimulation, and 39.2 ± 7.3 years in male partners. The biochemical pregnancy rate was 55.7%, the clinical pregnancy rate 46.3%, and the clinical miscarriage rate 14.6%. To identify predictors of clinical pregnancy loss, binary logistic regression with Akaike information criterion and multiple machine learning feature-selection methods (Boruta, recursive feature elimination, genetic algorithm, and LASSO regression) were applied. Six variables were consistently selected by at least two methods. Five were maternal-related: age, BMI, AMH levels, pre-transfer progesterone levels, and uterine abnormalities. Paternal age was the only male factor retained. No embryo-related variables (embryo quality, day of biopsy, mosaicism) or biopsy characteristics were selected. A predictive model using these six variables was developed. Multiple classification algorithms were evaluated, including neural networks, support vector machines, neighborhood-based methods, decision trees, gradient boosting, and bagging methods. Model optimization was performed by maximizing the AUC. Random Forest achieved the best performance, with an AUC of 85.7%, accuracy of 79.0%, positive predictive value of 76.5%, and negative predictive value of 82.1%. SHAP values were used to evaluate the influence of each variable on the model predictions. Limitations, reasons for caution The retrospective design limits causal inference. Prospective, multicentre studies are required to externally validate these predictors and confirm their clinical relevance. Wider implications of the findings Identification of non-chromosomal predictors of clinical pregnancy loss after PGT-A may improve patient counselling, refine prognostic assessment, and guide targeted interventions aimed at reducing miscarriage risk in IVF treatments. Trial registration number No
Zitieren
APA
Ortiz, J., Belloni, L., Lledó, B., Morales, R., Lozano, F., Rodríguez-Arnedo, A., & Bernabeu, A. (2026). L26/P-626 Factors affecting clinical pregnancy loss following euploid embryo transfer in PGT-A cycles: A machine learning approach. Human Reproduction, 41(Supplement_1). https://doi.org/10.1093/humrep/deag083.956
Vancouver
Ortiz J, Belloni L, Lledó B, Morales R, Lozano F, Rodríguez-Arnedo A, et al. L26/P-626 Factors affecting clinical pregnancy loss following euploid embryo transfer in PGT-A cycles: A machine learning approach. Human Reproduction. 2026;41(Supplement_1). doi:10.1093/humrep/deag083.956
BibTeX
@article{Ortiz2026,
author = {Ortiz, J.A. and Belloni, L. and Lledó, B. and Morales, R. and Lozano, F.M. and Rodríguez-Arnedo, A. and Bernabeu, A.},
title = {L26/P-626 Factors affecting clinical pregnancy loss following euploid embryo transfer in PGT-A cycles: A machine learning approach},
journal = {Human Reproduction},
year = {2026},
volume = {41},
number = {Supplement_1},
doi = {10.1093/humrep/deag083.956}
}
EUVIMED