Research on Decision-Making for NIPT Testing Timing Based on Mixed-Effects Models and Risk Optimization

Abstract

Noninvasive prenatal testing (NIPT), which analyzes fetal cell-free DNA in maternal blood to detect chromosomal aneuploidies early in pregnancy, faces challenges in balancing accuracy and timeliness due to the influence of individual variations such as gestational age and body mass index (BMI). To address this issue, this study proposes a dynamic modeling framework integrating clustering and machine learning to optimize the timing of NIPT and improve individualized adaptability. Using fetal Y-chromosome concentration as a core variable, we constructed individual growth curves and mixed-effects models to characterize interactions between gestational age and BMI, combined with Gaussian mixture modeling (GMM) for population stratification. By incorporating survival analysis and risk optimization functions, optimal testing windows were identified for different maternal subgroups, with model robustness verified via Monte Carlo simulations. Results demonstrate that the optimal testing window concentrates between 12–14 weeks of gestation, with the model exhibiting insensitivity to error perturbations. Further integration of multidimensional features—including maternal age, X chromosome concentration, and reproductive history—enabled feature importance analysis through random forest methods, establishing an interpretable high-dimensional clustering prediction framework. This study provides quantitative evidence to support intelligent decision-making in the scheduling and strategy of NIPT.