L26/P-828 Beyond a single leading follicle: follicle size distribution on trigger day reveals a meaningful window for mature oocyte yield across ovarian response categories
Human Reproduction · 01.07.2026 · 41 (Supplement_1)
DOI: 10.1093/humrep/deag083.1155 ↗Kurz gefasst
Fachartikel in der Zeitschrift Human Reproduction, online erschienen am 08.07.2026. Verfasst von X.L. Bui und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Study question Can follicle size distribution on trigger day predict mature oocyte yield and identify a consistent follicle size window across ovarian response categories? Summary answer A distribution-based assessment of follicle size enhances prediction of mature oocyte yield, consistently identifying a 13-18 mm follicle window across ovarian response subgroups. What is known already Follicle size on trigger day reflects endocrine-driven folliculogenesis and is closely associated with oocyte maturity. However, clinical assessment during ovarian stimulation often focuses on leading follicles or total follicle counts, which may overlook the contribution of the full follicle size distribution, particularly in patients with a limited follicular response where a dominant leading follicle may not be achieved. The existence of a biologically meaningful follicle size window driving mature oocyte yield across different ovarian response profiles remains insufficiently characterised. Conventional linear or count-based statistical approaches may inadequately capture the complex, non-linear relationship between follicle development and oocyte maturity. Study design, size, duration This retrospective cohort study analysed a clinical dataset of 4,472 antagonist Intracytoplasmic sperm injection (ICSI) cycles. After data preprocessing, including restriction to first ovarian stimulation cycles and exclusion of cycles with missing or non-eligible data, 907 cycles were included for final analysis. Participants/materials, setting, methods Trigger-day follicle counts were categorised into predefined size ranges (9-10, 11-12, 13-14, 15-16, 17-18 and 19-20 mm). The primary outcome was the number of mature (MII) oocytes retrieved. Negative binomial regression served as a statistical baseline and was compared with random forest and gradient boosting machine (GBM) models. Performance was evaluated using 5-fold out-of-fold cross-validation, calibration analysis and permutation-based feature importance. Subgroup analyses were performed according to ovarian response (low, normal and high). Main results and the role of chance Negative binomial regression demonstrated limited predictive performance (R² ∼ 0.28; MAE ∼ 3.0 MII) and systematic underestimation at higher mature oocyte counts. In contrast, machine learning models substantially improved prediction accuracy, with GBM achieving the best performance (R² ∼ 0.60; MAE ∼ 2.66 MII), closely followed by random forest. Both models showed good overall calibration across the range of predicted MII values, indicating reliable performance without systematic bias. Permutation-based feature importance consistently highlighted a mid-range follicle size window as the dominant contributor to mature oocyte prediction. Across the full cohort, follicles measuring 13-14 mm and 15-16 mm showed the greatest impact on model performance, followed by 17-18 mm follicles, whereas smaller (9-10 mm) and very large (19-20 mm) follicles contributed less. Subgroup analysis by ovarian response demonstrated robust and biologically consistent feature importance patterns. Across low, normal and high responders, follicles measuring 13-18 mm consistently ranked among the strongest contributors to mature oocyte prediction, each associated with the largest increases in prediction error when permuted, despite differences in overall response magnitude. These findings suggest that follicles within the 13-18 mm range reflect a coordinated maturation window rather than isolated dominant follicles effects. Limitations, reasons for caution The retrospective single-centre design may limit generalisability. Hormonal profiles and stimulation protocol variations were not included and may further refine prediction and biological interpretation in future studies. Wider implications of the findings These findings support a distribution-based assessment of follicular development during ovarian stimulation, identifying a biologically meaningful 13-18 mm follicle window that may inform more individualised trigger-day decision-making in IVF. Trial registration number No
Zitieren
APA
Bui, X., & Le, N. (2026). L26/P-828 Beyond a single leading follicle: follicle size distribution on trigger day reveals a meaningful window for mature oocyte yield across ovarian response categories. Human Reproduction, 41(Supplement_1). https://doi.org/10.1093/humrep/deag083.1155
Vancouver
Bui X, Le N. L26/P-828 Beyond a single leading follicle: follicle size distribution on trigger day reveals a meaningful window for mature oocyte yield across ovarian response categories. Human Reproduction. 2026;41(Supplement_1). doi:10.1093/humrep/deag083.1155
BibTeX
@article{Bui2026,
author = {Bui, X.L. and Le, N.Q.K.},
title = {L26/P-828 Beyond a single leading follicle: follicle size distribution on trigger day reveals a meaningful window for mature oocyte yield across ovarian response categories},
journal = {Human Reproduction},
year = {2026},
volume = {41},
number = {Supplement_1},
doi = {10.1093/humrep/deag083.1155}
}
EUVIMED