A machine learning algorithm can optimize the day of trigger to improve in vitro fertilization outcomes

Area of Science:

• Reproductive Medicine
• Artificial Intelligence in Healthcare
• Biostatistics

Background:

• Optimizing trigger injection timing is crucial for maximizing oocyte yield in in vitro fertilization (IVF).
• Physician-dependent decisions may not always achieve optimal outcomes.
• Machine learning offers potential for data-driven optimization in clinical decision-making.

Purpose of the Study:

• To evaluate if a machine learning causal inference model can optimize trigger injection timing.
• To maximize the number of fertilized oocytes (2PNs) and total usable blastocysts per IVF cycle.

Main Methods:

• A T-learner model with bagged decision trees was employed for causal inference.
• The model recommended trigger timing based on patient characteristics and stimulation parameters.
• A retrospective analysis of 7,866 IVF cycles from 2008-2019 was conducted.

Main Results:

• Algorithm-assisted decisions resulted in an average of 1.430 more 2PNs and 0.577 more usable blastocysts per stimulation compared to physician decisions.
• Following the model's recommendation increased 2PNs by 3.015 and usable blastocysts by 1.515 on average.
• Key predictive features included follicle count (16-20mm and 11-15mm) and estradiol levels.

Conclusions:

• Machine learning-driven optimization of trigger timing can significantly enhance IVF cycle outcomes.
• This AI approach shows potential to improve the efficiency and success rates of fertility treatments.
• Prospective studies are needed to validate these findings in clinical practice.