An integrated multi-factor decision model for personalized NIPT timing using BMI stratification

Abstract Non-invasive prenatal testing (NIPT) is a prenatal screening technique that analyzes cell-free fetal DNA in maternal blood plasma, widely used for early detection of fetal chromosomal abnormalities such as trisomy 21, 18, and 13. To optimize personalized testing timing and improve the detection accuracy of chromosomal abnormalities in female fetuses, this study developed a data-driven multifactorial modeling framework. The study population consisted of 1,687 pregnant women from a certain region. Correlation analysis and a nonlinear mixed-effects model (M4) were employed to quantify the relationships among gestational age, body mass index (BMI), and Y-chromosome concentration. The model results showed a positive correlation between gestational age and Y-chromosome concentration, while BMI and height were significantly negatively correlated with Y-chromosome concentration, providing statistical support for subsequent BMI-based stratification. Building on this, the study further predicted the earliest possible gestational week at which fetal DNA concentration reaches the detection threshold for each pregnant woman. A joint optimization model for BMI grouping and testing timing was constructed to determine the optimal testing time for each group, and the interaction between BMI and gestational age was validated using K-means clustering incorporating multiple factors. The model results indicated that the M4 model exhibited the best goodness of fit. As BMI increased, the recommended testing time was postponed from 13.8 weeks to 21.5 weeks. Additionally, to address the issues of limited data and the absence of Y-chromosome signals in female fetuses, this study systematically compared decision tree, random forest, and support vector machine (SVM) models, constructing a multi-feature fusion-based classification system. In the task of classifying chromosomal abnormalities in female fetuses, the SVM model performed the best, with X-chromosome concentration and GC content identified as key discriminatory features. The comprehensive multifactorial model proposed in this study can effectively predict personalized NIPT timing, help reduce the risk of testing failure, enhance the detection performance for chromosomal abnormalities in female fetuses, and thereby provide scientific basis and methodological support for the clinical application of NIPT.