To develop and internally validate a machine learning model for predicting the risk of large-for-gestational-age (LGA) birth following IVF/ICSI and to identify important parental and treatment-related predictors.
Approach:
Data Collection: Data on 17,741 singleton live births resulting from IVF/ICSI were collected, categorized as appropriate for gestational age (AGA) or LGA.
Model Development: An XGBoost model was developed and optimized using the Optuna framework and Tree-structured Parzen Estimator algorithm.
Model Evaluation: Model performance was evaluated using AUC, calibration analysis, Brier score, and classification metrics; logistic regression served as a baseline model.
Interpretability Analysis: Model interpretability was assessed using SHAP analysis.
Key Findings:
The XGBoost model achieved an AUC of 0.7003 in the internal hold-out test set, compared to 0.6445 for logistic regression.
Calibration analysis showed a lower Brier score for XGBoost (0.2040) than for logistic regression (0.2295).
Important predictors identified included embryo transfer strategy, maternal anthropometric and metabolic characteristics, and several paternal characteristics.
Interpretation:
The machine learning model demonstrated moderate discriminative ability for predicting LGA risk among IVF/ICSI-conceived singleton births in internal validation.
Limitations:
The model's clinical utility requires future prospective external validation.
Conclusion:
The findings indicate a significant association of embryo cryopreservation strategies, maternal and paternal factors with fetal overgrowth.
