Revisiting the Trigger-to-Oocyte Retrieval Interval: A Machine Learning-Guided Strategy for Improving Oocyte Yield in...

Area of Science:

• Reproductive Medicine
• In Vitro Fertilization (IVF)
• Ovarian Reserve

Background:

• Standard IVF protocols use a trigger-to-oocyte retrieval interval (TOI) of 34-36 hours to prevent premature ovulation.
• Women with diminished ovarian reserve (DOR) or advanced age often experience premature ovulation or empty follicles with standard TOI.
• Nonsteroidal anti-inflammatory drugs (NSAIDs) can prolong the ovulatory window, suggesting potential for TOI extension.

Purpose of the Study:

• To evaluate the impact of extending the TOI (≥36 hours) on mature oocyte rate (MOR) and oocyte retrieval rate (ORR).
• To develop and validate machine learning models for predicting oocyte yield and optimizing individualized TOI in IVF.
• To investigate TOI optimization strategies for patients with diminished ovarian reserve (DOR).

Main Methods:

• Retrospective analysis of 15,033 IVF cycles to assess the effects of extended TOI (≥36 hours) on MOR and ORR.
• Training seven machine learning models to predict mature oocytes (MII) and retrieved oocytes (Oocount).
• External validation of the best-performing model on 387 unseen cases to assess predictive accuracy for MOR and ORR.

Main Results:

• Extended TOI significantly improved MOR (p=0.00034) and ORR (p=0.0051) in POSEIDON groups 3 and 4, particularly benefiting DOR patients.
• The Extreme Gradient Boosting (XGBoost) model demonstrated the highest predictive accuracy for MII and Oocount.
• External validation showed a decrease in R-squared but maintained acceptable MAE and RMSE, indicating model reliability.

Conclusions:

• Extending the TOI is a viable strategy to enhance both MOR and ORR in patients with DOR.
• Machine learning models can accurately predict oocyte yield, enabling calculation of MOR and ORR across various TOIs.
• Individualized TOI optimization using predictive models facilitates improved IVF outcomes and retrieval timing.