L26/P-223 Key predictors of usable blastocyst yield in highly purified human menopausal gonadotropin (hp-hMG) only cycles: a multicenter machine learning analysis
Human Reproduction · 01.07.2026 · 41 (Supplement_1)
DOI: 10.1093/humrep/deag083.559 ↗Kurz gefasst
Fachartikel in der Zeitschrift Human Reproduction, online erschienen am 08.07.2026. Verfasst von N. Itzhak und 13 weiteren Autor:innen.
Abstract (Originalsprache)
Study question Which clinical and laboratory factors are the strongest predictors of usable blastocyst yield in hp-hMG-only IVF/ICSI cycles? Summary answer Fertilization (2PN) count, maternal age, and oocyte yield were the dominant predictors. The model achieved high accuracy (R2=0.55, MAE=1.23) in independent testing. What is known already Estimating the number of usable blastocysts is critical for patient counseling and laboratory planning. However, clinicians typically rely on imperfect proxies like age, ovarian reserve markers, or simple oocyte counts to guess this outcome. These metrics often fail to capture the complex attrition from retrieval to Day 5. Existing AI tools predominantly focus on predicting oocyte yield or pregnancy outcomes. Evidence quantifying usable blastocyst yield remains limited, leaving a gap in tools that provide precise, quantitative yield expectations. Study design, size, duration This retrospective multicenter cohort study analyzed routine clinical and laboratory data from IVF clinics in Europe, North America, and Asia (2014–2024). hp-hMG-only cycles with complete records for trigger parameters and oocyte outcomes were included. The model was developed on 974 cycles. Validation was performed on an independent test set of 345 cycles with complete fertilization and usable blastocyst outcomes. The primary endpoint was the count of usable blastocysts. Participants/materials, setting, methods Inputs included age, BMI, AMH, AFC, trigger-day hormones (E2, P4), follicle counts, starting dose, and retrieved oocytes (Total/MII). An XGBoost regression model predicted usable blastocyst count using these features plus fertilization (2PN) data. Feature importance was analyzed to rank variables by their contribution to predictive accuracy. Performance was evaluated on the independent 345-cycle test set. Main results and the role of chance In the independent test set (n = 345), the mean yield was 2.56 usable blastocysts. The model demonstrated reliable predictive performance of blastocysts number with R2=0.55 and MAE=1.23. Feature importance analysis singled out fertilization count (2PN) as the dominant predictor. However, the model relied on additional clinical factors to refine accuracy beyond simple fertilization numbers. The next two most impactful features were maternal age and mature (MII) oocyte number, which contributed significantly to the prediction even when 2PN was known. Other impactful variables in descending order were total oocytes retrieved, estradiol level at trigger day, and Antral Follicle Count (AFC). Suggesting that final blastocyst yield is not determined by fertilization success alone but is substantially modulated by patient age, ovarian response and treatment characteristics. BMI’s impact was not considered high by the model, possibly due to its impact being mitigated by personalized dose adjustments during stimulation. Limitations, reasons for caution Limitations include the retrospective design and restriction to hp-hMG -only cycles, which limits generalizability to other stimulation protocols. Unmeasured variations in laboratory culture and grading criteria may affect results. Prospective external validation is required before clinical implementation. Wider implications of the findings Accurate blastocyst prediction requires integrating fertilization data with age, ovarian markers and treatment parameters. AI models can assist us in improving our understanding of each parameter’s relative contribution to the blastulation rate, improving prediction capabilities and possibly fine-tuning future treatments for better outcomes. Trial registration number No
Zitieren
APA
Itzhak, N., Luz, A., Hourvitz, R., Youngster, M., Hochberg, A., Baum, M., Maman, E., Arie, D. C. B., Yu, M., Zhou, W., Barletta, K., Hariton, E., Polyzos, N., & Hourvitz, A. (2026). L26/P-223 Key predictors of usable blastocyst yield in highly purified human menopausal gonadotropin (hp-hMG) only cycles: a multicenter machine learning analysis. Human Reproduction, 41(Supplement_1). https://doi.org/10.1093/humrep/deag083.559
Vancouver
Itzhak N, Luz A, Hourvitz R, Youngster M, Hochberg A, Baum M, et al. L26/P-223 Key predictors of usable blastocyst yield in highly purified human menopausal gonadotropin (hp-hMG) only cycles: a multicenter machine learning analysis. Human Reproduction. 2026;41(Supplement_1). doi:10.1093/humrep/deag083.559
BibTeX
@article{Itzhak2026,
author = {Itzhak, N. and Luz, A. and Hourvitz, R. and Youngster, M. and Hochberg, A. and Baum, M. and Maman, E. and Arie, D. Chuderland Ben and Yu, M. and Zhou, W. and Barletta, K. and Hariton, E. and Polyzos, N.P. and Hourvitz, A.},
title = {L26/P-223 Key predictors of usable blastocyst yield in highly purified human menopausal gonadotropin (hp-hMG) only cycles: a multicenter machine learning analysis},
journal = {Human Reproduction},
year = {2026},
volume = {41},
number = {Supplement_1},
doi = {10.1093/humrep/deag083.559}
}
EUVIMED