Developing a live birth prediction model for single euploid frozen embryo transfer using machine learning: a novel approach

Dear Editor,

I read the article entitled “Predictive model for live birth outcomes in single euploid frozen embryo transfers: a comparative analysis of logistic regression and machine learning approaches” [1] and commend its focus on a critical yet often overlooked aspect of single euploid frozen embryo transfer (seFET)—the construction of a live birth (LB) prediction model. This study developed a comprehensive predictive model integrating patient characteristics (e.g., BMI, AMH, infertility type, and endometrial thickness) and embryo parameters (e.g., blastocyst quality and trophectoderm biopsy day) to predict LB outcomes in seFET cycles. While previous studies have primarily focused on the impact of embryo morphology or single variables on IVF outcomes, this research fills a gap by considering multiple factors in predicting LB outcomes in seFET cycles.

The study utilized data from 1979 seFET cycles spanning 2017 to 2023, offering a large sample size and extended timeframe, which enhances the reliability and generalizability of the findings. Furthermore, rigorous statistical analyses were conducted, including nearly exhaustive repeated cross-validation (60:40) and fourfold cross-validation, ensuring model stability and calibration performance. The study also implemented strict controls and evaluations to address potential overfitting, further reinforcing the credibility and applicability of the results.

One of the most striking findings of this research is the innovative development of predictive models. In addition to traditional logistic regression (LR), the study compared multiple machine learning (ML) methods, such as random forest classifier (RFC), support vector machine (SVM), and gradient boosting (XGBoost). As an emerging computational tool in the twenty-first century, ML holds immense potential in healthcare, particularly in predicting binary disease outcomes. By comparing various ML algorithms, the study demonstrated that factors such as BMI, blastocyst quality, and TE biopsy day significantly influence LB outcomes. These findings provide valuable guidance for clinicians in embryo selection and patient counseling.

To enhance the clinical utility of the research, the authors also developed a freely accessible online prediction tool (ABLE), enabling clinicians and patients to obtain real-time LB probability estimates based on individual characteristics and embryo parameters. This tool serves as a powerful resource for patient counseling and decision-making, facilitating the global dissemination of the study’s findings and assisting physicians in evaluating patient outcomes.

The article also highlights that, although the model performed well in internal validation, further external validation across diverse geographical and clinical settings is necessary to confirm its generalizability and robustness. Future studies are still needed to address remaining gaps in this field.

In summary, this article provides valuable insights into the key factors influencing LB outcomes in single euploid frozen embryo transfer. Notably, it innovatively applies ML methods in model construction, suggesting broader applications of ML in reproductive medicine for predictive modeling. This approach not only meets patients’ needs for accurate LB success rate counseling but also assists clinicians in making evidence-based decisions. By doing so, we can potentially improve the quality of healthcare worldwide and offer a brighter future for individuals undergoing in vitro fertilization.

Author contribution

TT: ideas, writing, analyzing, and approval.

Declarations

Conflict of interest

The author declares no competing interests.