Skip to content
DraconiticResearch
Search Sources API
203,234 studies31 sourcesUpdated 37 min ago
Peer-reviewedCC BY 4.0BMC Pregnancy and Childbirth

Estimating individualized gonadotropin starting doses in women with PCOS: a target-driven machine learning model for GnRH-antagonist stimulation

  • Wenjia Meng1
  • Ye Pan1
  • Moxuan Zhang2
  • Ziyi Song1,3
  • Zerui Xiao3
  • Rong Liang3
  • Xiao Shi1
  • Hongchu Bao4
  • Lili Zhuang4
  • Qun Lu1,3
4 affiliations
  1. 1Medical Center for Human Reproduction, Beijing Chao-Yang Hospital, Capital Medical University, Beijing, 100020 China
  2. 2Beijing Neurosurgical Institute, Beijing Tiantan Hospital, Capital Medical University, Beijing, 100070 China
  3. 3Center of Reproductive Medicine, Peking University People’s Hospital, Beijing, 100044 China
  4. 4Reproductive Medicine Centre, Yuhuangding Hospital of Yantai, Affiliated Hospital of Qingdao University, Yantai, 264000 China
Read full textDOICite

Abstract

Background: Women with polycystic ovary syndrome (PCOS) are prone to excessive ovarian response and ovarian hyperstimulation syndrome (OHSS) during controlled ovarian stimulation, whereas overly conservative gonadotropin (Gn) dosing reduces oocyte yield and efficiency. This study aimed to develop a target-driven machine learning model for estimating individualized Gn starting doses in women with PCOS.

Methods: This retrospective, two-center study analyzed 976 PCOS women undergoing in vitro fertilization-embryo transfer using GnRH-antagonist protocols. Patients were categorized by oocyte yield (< 7, 7–16, > 16). Ovarian sensitivity (OS), defined as retrieved oocytes per unit of Gn starting dose, was modeled as a continuous response metric linking baseline characteristics with Gn requirements. Data were partitioned into training, validation, and test sets (60:20:20). Parallel random forest (RF) and least absolute shrinkage and selection operator (LASSO) regression identified core predictors within the training set. Model accuracy was evaluated via R2, RMSE, and MAE.

Results: The 7–16 oocyte-yield category showed a more favorable balance of reproductive outcomes, OHSS risk, time to live birth, and costs. Body mass index (BMI), anti-Müllerian hormone (AMH), basal follicle-stimulating hormone (bFSH), and antral follicle count (AFC) were retained as final predictors. The RF model robustly predicted continuous OS in the independent test set, achieving an R2 of 0.860 (95% CI: 0.809, 0.901), an RMSE of 0.254 (95% CI: 0.218, 0.296), and an MAE of 0.141 (95% CI: 0.127, 0.156). Post-hoc dose-concordance analysis in the 7–16 category showed model-derived doses correlated with historical prescriptions. Simulated dose-adjustment analyses generated upward recommendations for < 7 oocytes and downward recommendations for > 16 oocytes. An online “Gn Calculator” was developed for clinical decision support.

Conclusions: By predicting observed OS via BMI, AMH, bFSH, and AFC, this target-driven RF framework calculates individualized Gn starting doses for women with PCOS. The Gn Calculator facilitates clinical application and may help guide patients toward optimal outcomes by targeting an ideal oocyte yield.

Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1186/s12884-026-09783-x.

Related

  1. Real-world implementation of AI-assisted starting gonadotropin dosing in routine ovarian stimulation: an exposure-defined, propensity score-matched observational cohort study
    Huiyu Xu, Congcong Ma, Guoshuang Feng +7
    Frontiers in EndocrinologyPeer-reviewedCC BY 4.0Endocrinology & metabolismCrossref
  2. Estimating Individualized Survival Benefit of Consolidative Metastasis-Directed Radiotherapy Following Chemoimmunotherapy in Nasopharyngeal Carcinoma With Exclusive Distant Metastasis.
    Xuguang Wang, Shuwei Chen, Xing Zhang +11
    MedCommPeer-reviewedCC BYFull textGeneral medicineEurope PMC
  3. Comparative effectiveness of GnRH antagonist, mild stimulation, and PPOS protocols in POSEIDON groups 3 and 4: a retrospective cohort study with entropy balancing
    Haiyue Zhao, Yunping Ni, Aiai Wang +3
    Frontiers in EndocrinologyPeer-reviewedCC BY 4.0Full textEpidemiology & public healthCrossrefPMC
  4. Adverse effects of a superovulatory regimen of equine chorionic gonadotropin and gonadotropin-releasing hormone on embryo implantation in mice
    Nana Yang, Jingmei Han, Xueyan Wang +2
    Veterinary Research ForumPeer-reviewedCC BY-NC-SA 4.0Full textObstetrics, gynaecology & paediatricsPMC
  5. Undifferentiated pleomorphic sarcoma expressing human chorionic gonadotropin beta subunit.
    Michał Piekarz, Anna Szczerba, Aleksandra Śliwa +4
    Reports of practical oncology and radiotherapy : journal of Greatpoland Cancer Center in Poznan and Polish Society of Radiation OncologyPeer-reviewedCC BY-NC-NDFull textOncologyEurope PMC

Cite

BibTeXRIS
BibTeX
@article{meng2026estimating,
  title = {{Estimating individualized gonadotropin starting doses in women with PCOS: a target-driven machine learning model for GnRH-antagonist stimulation}},
  author = {Wenjia Meng and Ye Pan and Moxuan Zhang and Ziyi Song and Zerui Xiao and Rong Liang and Xiao Shi and Hongchu Bao and Lili Zhuang and Qun Lu},
  journal = {BMC Pregnancy and Childbirth},
  year = {2026},
  month = aug,
  doi = {10.1186/s12884-026-09783-x},
  url = {https://doi.org/10.1186/s12884-026-09783-x},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Estimating individualized gonadotropin starting doses in women with PCOS: a target-driven machine learning model for GnRH-antagonist stimulation
AU  - Wenjia Meng
AU  - Ye Pan
AU  - Moxuan Zhang
AU  - Ziyi Song
AU  - Zerui Xiao
AU  - Rong Liang
AU  - Xiao Shi
AU  - Hongchu Bao
AU  - Lili Zhuang
AU  - Qun Lu
PY  - 2026
DA  - 2026/08/10
JO  - BMC Pregnancy and Childbirth
DO  - 10.1186/s12884-026-09783-x
UR  - https://doi.org/10.1186/s12884-026-09783-x
AB  - Background: Women with polycystic ovary syndrome (PCOS) are prone to excessive ovarian response and ovarian hyperstimulation syndrome (OHSS) during controlled ovarian stimulation, whereas overly conservative gonadotropin (Gn) dosing reduces oocyte yield and efficiency. This study aimed to develop a target-driven machine learning model for estimating individualized Gn starting doses in women with PCOS.  Methods: This retrospective, two-center study analyzed 976 PCOS women undergoing in vitro fertilization-embryo transfer using GnRH-antagonist protocols. Patients were categorized by oocyte yield (< 7, 7–16, > 16). Ovarian sensitivity (OS), defined as retrieved oocytes per unit of Gn starting dose, was modeled as a continuous response metric linking baseline characteristics with Gn requirements. Data were partitioned into training, validation, and test sets (60:20:20). Parallel random forest (RF) and least absolute shrinkage and selection operator (LASSO) regression identified core predictors within the training set. Model accuracy was evaluated via R2, RMSE, and MAE.  Results: The 7–16 oocyte-yield category showed a more favorable balance of reproductive outcomes, OHSS risk, time to live birth, and costs. Body mass index (BMI), anti-Müllerian hormone (AMH), basal follicle-stimulating hormone (bFSH), and antral follicle count (AFC) were retained as final predictors. The RF model robustly predicted continuous OS in the independent test set, achieving an R2 of 0.860 (95% CI: 0.809, 0.901), an RMSE of 0.254 (95% CI: 0.218, 0.296), and an MAE of 0.141 (95% CI: 0.127, 0.156). Post-hoc dose-concordance analysis in the 7–16 category showed model-derived doses correlated with historical prescriptions. Simulated dose-adjustment analyses generated upward recommendations for < 7 oocytes and downward recommendations for > 16 oocytes. An online “Gn Calculator” was developed for clinical decision support.  Conclusions: By predicting observed OS via BMI, AMH, bFSH, and AFC, this target-driven RF framework calculates individualized Gn starting doses for women with PCOS. The Gn Calculator facilitates clinical application and may help guide patients toward optimal outcomes by targeting an ideal oocyte yield.  Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1186/s12884-026-09783-x.
SN  - 1471-2393
ER  -