EUVIMEDEUVIMED entdecken
Crossref

L26/P-751 Development of a personalized dose-response curve for initial r-hFSH dose selection in ovarian stimulation cycles

A. Hourvitz, N. Itzhak, A. Luz, R. Hourvitz, E. Moran, S. Reuvenny, A. Hochberg, M. Baum, E. Hariton, N.P. Polyzos, W. Zheng, T. D’Hooghe und 1 weitere

Human Reproduction · 01.07.2026 · 41 (Supplement_1)

DOI: 10.1093/humrep/deag083.1078 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Human Reproduction, online erschienen am 08.07.2026. Verfasst von A. Hourvitz und 12 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Study question Can a machine learning framework generate a clinically robust, personalized r-hFSH dose-response curve predicting oocyte yield across the full dose range? Summary answer A patient-specific sigmoid dose-response curve was developed, effectively visualizing the effect of varying initial r-hFSH dose on oocyte yield. What is known already The FSH starting dose-oocyte-yield relationship remains controversial. Initial dosing relies on subjective clinician experience or generalized nomograms. Nomograms focus on achieving a specific oocyte number, rather than providing a full range of oocyte-yields according to varying FSH starting doses. The few previously published dose-response models insufficiently addressed dose selection bias; they incorrectly inferred that high doses cause poorer outcomes, overlooking the causal relationship where clinicians prescribe high FSH starting doses to low-prognosis patients. A bias-aware tool that can generate patient-specific dose–response curves isolating the true effect of FSH dosing for defined patient characteristics hasn’t been developed yet. Study design, size, duration This retrospective, multi-center cohort study collected data from 10 IVF clinics across Europe, North America, and Asia (2014–2024), utilizing patient parameters including Age, BMI, AMH, and AFC. The dataset consisted of 22,686 mixed-protocol cycles used for model development and 778 r-hFSH-only cycles for testing. The study focused on constructing and validating a model that generates personalized dose-response curves for ovarian stimulation with r-hFSH, ensuring the output remains accurate and clinically sound across diverse patient profiles. Participants/materials, setting, methods A multi-step modeling framework was developed. First, an outlier detection model defined a valid dose range using patient parameters, avoiding rare scenarios where data is sparse to ensure predictions reliability. Next, an XGBoost-based algorithm modeled the effect of dose on outcome, accounting for the causal relationship between patient parameters and dose. Finally, predictions were fitted to a sigmoid function, generating a continuous curve modeling the plateau in ovarian response capped by the finite ovarian reserve. Main results and the role of chance The model demonstrated strong predictive performance, significantly exceeding the accuracy reported in previous publications. The overall model achieved an R2 of 0.66 with a Mean Absolute Error (MAE) of just 3.65 oocytes across the entire cohort (n = 778). To assess reliability across different patient types, we also analyzed performance by ovarian reserve levels (AFC) for all cycles that had available AFC information (n = 644): The prediction error increased along with the AFC in the subgroups (from 2.19 in poor responders to 4.24 in hyper-responders). However, this variation was driven by the parallel increase in mean oocyte yield rather than model instability. The error is proportional to the average oocyte yield, while remaining relatively low even in hyper-responders (AFC > 25), confirming the tool is robust for both low and high-prognosis patients. Limitations, reasons for caution The primary limitation is the study’s retrospective design. The model also does not account for key clinical factors such as previous cycle ovarian response, reasons for specific treatment, or infertility causes. These variables were deliberately omitted to prioritize the tool’s ease of use. Wider implications of the findings This model offers clinicians a robust, data-driven tool for r-hFSH starting dose optimization, enhancing personalization in ovarian stimulation cycles. Its performance across diverse patient profiles can improve outcomes and optimize resource use, supporting the broader adoption of ML in reproductive medicine. Trial registration number Yes

Zitieren

APA

Hourvitz, A., Itzhak, N., Luz, A., Hourvitz, R., Moran, E., Reuvenny, S., Hochberg, A., Baum, M., Hariton, E., Polyzos, N., Zheng, W., D’Hooghe, T., & Maman, E. (2026). L26/P-751 Development of a personalized dose-response curve for initial r-hFSH dose selection in ovarian stimulation cycles. Human Reproduction, 41(Supplement_1). https://doi.org/10.1093/humrep/deag083.1078

Vancouver

Hourvitz A, Itzhak N, Luz A, Hourvitz R, Moran E, Reuvenny S, et al. L26/P-751 Development of a personalized dose-response curve for initial r-hFSH dose selection in ovarian stimulation cycles. Human Reproduction. 2026;41(Supplement_1). doi:10.1093/humrep/deag083.1078

BibTeX
@article{Hourvitz2026,
  author = {Hourvitz, A. and Itzhak, N. and Luz, A. and Hourvitz, R. and Moran, E. and Reuvenny, S. and Hochberg, A. and Baum, M. and Hariton, E. and Polyzos, N.P. and Zheng, W. and D’Hooghe, T. and Maman, E.},
  title = {L26/P-751 Development of a personalized dose-response curve for initial r-hFSH dose selection in ovarian stimulation cycles},
  journal = {Human Reproduction},
  year = {2026},
  volume = {41},
  number = {Supplement_1},
  doi = {10.1093/humrep/deag083.1078}
}
Kostenlos starten →