L26/P-269 Stage-Structured, Distributional Prediction of IVF Outcomes with Conditional Updating
Human Reproduction · 01.07.2026 · 41 (Supplement_1)
DOI: 10.1093/humrep/deag083.605 ↗Kurz gefasst
Fachartikel in der Zeitschrift Human Reproduction, online erschienen am 08.07.2026. Verfasst von A. Craig und 5 weiteren Autor:innen.
Abstract (Originalsprache)
Study question Can a distribution-based, stage-structured model improve IVF outcome prediction by quantifying uncertainty and dynamically updating predictions as intermediate clinical outcomes become known? Summary answer A sequential distributional framework predicts full outcome distributions throughout IVF and updates downstream expectations when observed outcomes are entered, reducing uncertainty and improving clinical interpretability. What is known already Existing IVF prediction tools primarily provide static point estimates or cumulative success probabilities. These approaches fail to quantify uncertainty, do not reflect the sequential biology of IVF, and cannot incorporate mid-cycle information such as egg retrieval or embryo development outcomes, limiting their clinical usefulness for counseling and multi-cycle planning. Study design, size, duration Observational modeling study using UK national registry data from 2017–2018 combined with published clinical cohorts. The analysis included 103,924 fresh autologous IVF cycles for upstream stages and 429,507 additional observations across downstream stages including blastocyst formation, euploidy, cryosurvival, and live birth. Participants/materials, setting, methods De-identified cycle-level data from the HFEA were used to model egg retrieval, oocyte maturity, and fertilization. Downstream transitions were modeled using published international cohorts and registry summaries. Egg retrieval was modeled using a zero-inflated negative binomial specification, with subsequent stages propagated via stage-specific binomial filters. The framework maintains full probability distributions at each stage and implements conditional updating when observed outcomes become known. Multi-cycle scenarios were handled through distribution convolution while preserving cycle-specific parameters. Main results and the role of chance The framework produced full probability distributions for outcomes at each stage of IVF rather than single point estimates, enabling explicit quantification of biological uncertainty throughout treatment. Model validation demonstrated stable performance on held-out data with no evidence of meaningful overfitting, and predicted distributions were well calibrated across stages. Incorporation of observed intermediate outcomes, such as egg retrieval counts, led to intuitive and clinically coherent updates of downstream predictions, with uncertainty narrowing as additional information became available. This dynamic updating preserved patient and cycle-specific biology rather than averaging across populations, supporting more personalized counseling and decision-making. Limitations, reasons for caution The model relies on registry data with limited clinical covariates and grouped outcome reporting, requiring midpoint imputation. Some downstream stages depend on external published cohorts rather than unified individual-level data. Independence across cycles is assumed for multi-cycle convolution. Wider implications of the findings This distribution-based, dynamically updating framework addresses key limitations of existing IVF calculators by providing uncertainty-aware, stage-specific predictions that evolve with patient-specific outcomes. Such tools may improve shared decision-making, patient counseling, and modern family-planning strategies involving multi-cycle and elective IVF. Trial registration number No
Zitieren
APA
Craig, A., Wartschinski, L., Eyre, M., Davidson, I., Christensen, M., & Wolfram, T. (2026). L26/P-269 Stage-Structured, Distributional Prediction of IVF Outcomes with Conditional Updating. Human Reproduction, 41(Supplement_1). https://doi.org/10.1093/humrep/deag083.605
Vancouver
Craig A, Wartschinski L, Eyre M, Davidson I, Christensen M, Wolfram T. L26/P-269 Stage-Structured, Distributional Prediction of IVF Outcomes with Conditional Updating. Human Reproduction. 2026;41(Supplement_1). doi:10.1093/humrep/deag083.605
BibTeX
@article{Craig2026,
author = {Craig, A. and Wartschinski, L. and Eyre, M. and Davidson, I. and Christensen, M. and Wolfram, T.},
title = {L26/P-269 Stage-Structured, Distributional Prediction of IVF Outcomes with Conditional Updating},
journal = {Human Reproduction},
year = {2026},
volume = {41},
number = {Supplement_1},
doi = {10.1093/humrep/deag083.605}
}
EUVIMED