Individual predictions based on nonlinear mixed modeling: application to prenatal twin growth

Area of Science:

• Pediatrics
• Biostatistics
• Maternal-Fetal Medicine

Background:

• Current growth assessment relies on cross-sectional data, limiting individual trajectory analysis.
• Twin pregnancies face risks of growth restriction, necessitating improved methods for monitoring growth trajectories.
• Individual sequential predictions offer a more dynamic approach to assessing growth processes over time.

Purpose of the Study:

• To present a simple forecasting method for individual growth prediction.
• To improve the detection of abnormal growth trajectories, particularly in high-risk pregnancies like twins.
• To validate a novel prediction estimator using real-world twin pregnancy data.

Main Methods:

• Developed a forecasting method assuming individual growth follows population behavior.
• Utilized nonlinear mixed modeling to establish population parameters from a reference dataset.
• Proposed a prediction estimator without linear approximation, analyzed its asymptotic properties.
• Employed two independent real datasets of twin pregnancies for illustration and validation.

Main Results:

• The proposed method provides individual sequential predictions for growth assessment.
• The forecasting method was illustrated and validated on twin pregnancy datasets.
• The prediction estimator demonstrated favorable asymptotic behavior with increasing observations.

Conclusions:

• The developed forecasting method offers a valuable tool for assessing prenatal growth, especially in twin pregnancies.
• This approach enhances the ability to identify abnormal growth trajectories earlier and more accurately.
• The method's foundation in population modeling and validated performance supports its clinical utility.