Recent Advances in IVF and Reproductive Medicine

Over the past four weeks, several studies have highlighted significant advancements in IVF and reproductive medicine, offering new insights and potential improvements for fertility treatments. Below, we explore key findings that could shape future practices.

Improved IVF Predictions with Machine Learning

Yao MWM, Nguyen ET, Retzloff MG, Gago LA, Nichols JE, Payne JF, Ripps BA, Opsahl M, Groll J, Beesley R, Neal G, Adams J, Nowak L, Swanson T, Chen X. Machine learning center-specific models show improved IVF live birth predictions over US national registry-based model. Nat Commun. 2025 Apr 17

A notable study published on April 17, 2025, in Nature Communications Machine learning center-specific models show improved IVF live birth predictions compared machine learning center-specific (MLCS) models with the Society for Assisted Reproductive Technology (SART) national registry-based model for predicting live birth rates in IVF. The retrospective cohort study involved 4635 patients from six small-to-midsize US fertility centers across 22 locations in nine states and four regions (West, Southeast, Southwest, Midwest). The study design included center-specific test sets to evaluate model performance metrics and clinical utility.

Key findings include:

• MLCS models showed improved ROC-AUC compared to Age models (MLCS1: Z = 0.0, p < 0.05; MLCS2: Z = 0.0, p < 0.05).
• MLCS1 and MLCS2 had positive PLORA values, indicating enhanced predictive power over Age models.
• MLCS2 versus MLCS1 showed similar ROC-AUC (Z = 7.0, p > 0.05) but improved PLORA (MLCS2 median 23.9, IQR 10.2-39.4 vs. MLCS1 median 7.2, IQR 3.6-11.8, Z = 0.0, p < 0.05).
• External validation (LMV) of MLCS1 showed no significant difference from cross-validation in ROC-AUC and PLORA (ROC-AUC: Z = 5.0, p > 0.05; PLORA: Z = 6.0, p > 0.05).
• Data drift analysis revealed 1-2 clinical factors at two centers, 5-6 at three centers, and 8 at one center showed significant differences (p < 0.05), but this did not affect clinical utility.
• Compared to SART, MLCS2 had better Brier scores (Z = 0.0, p < 0.05), higher F1 scores at the 50% live birth probability (LBP) threshold (MLCS2 median 0.74, IQR 0.72-0.78 vs. SART median 0.71, IQR 0.68-0.73, Z = 0.0, p < 0.05), and higher PR-AUC (MLCS2 median 0.75, IQR 0.73-0.77 vs. SART median 0.69, IQR 0.68-0.71, Z = 0.0, p < 0.05).
• Reclassification showed 70% (3259/4645) concordant LBPs, with 30% (1386/4645) discordant, where MLCS2 assigned 26% (1230/4645) to higher LBP categories and 3.4% (156/4645) to lower compared to SART.

Clinical implications include enabling transparent, personalized IVF counseling and supporting value-based pricing, qualifying over 50% of patients. This contrasts with SART’s online calculator disclaimer, which notes, “The estimates are based on the data we have available and may not be representative of your specific experience…Please speak with your doctor about your specific treatment plan and potential for success.” This finding could revolutionize patient-centered care by addressing local data concerns, unlike the broader SART model.

Source: Machine learning center-specific models show improved IVF live birth predictions

Guinea Pig Embryos and Infertility Research

Canizo JR, Zhao C, Petropoulos S. The Guinea Pig Serves as an Alternative Model to Study Human Preimplantation Development. Nature Cell Biology. 2025;27(4):696-710. doi:10.1038/s41556-025-01642-9.

The Guinea Pig Serves as an Alternative Model to Study Human Preimplantation Development” published in Nature Cell Biology in 2025, provides a comprehensive analysis of guinea pig preimplantation development and its relevance to human embryogenesis.

The study utilized single-cell RNA sequencing to create an atlas of guinea pig preimplantation development. The findings revealed that guinea pig embryogenesis closely mirrors human preimplantation development in several key aspects, including the timing of compaction, blastocyst formation, and implantation. The spatio-temporal expression of key lineage markers was also found to be similar between guinea pigs and humans.

The research highlighted the conserved roles of critical signaling pathways such as Hippo, MEK-ERK, and JAK-STAT during preimplantation development. Additionally, a multi-species analysis was conducted, which underscored both conserved and divergent gene expressions during preimplantation development and pluripotency.

The study concluded that guinea pigs serve as a valuable model for advancing research in preimplantation development and stem cell biology. This model can be leveraged to better understand the long-term impacts of early exposures on offspring outcomes, providing a robust platform for translational research in reproductive medicine.

This manuscript underscores the potential of guinea pigs as an alternative model to study human embryogenesis, offering insights that could enhance the understanding and improvement of assisted reproductive technologies.

Source: https://pubmed.ncbi.nlm.nih.gov/40185949/