Author(s)

Zhiyu Wang

Corresponding Author

Zhiyu Wang

Abstract

Non-invasive prenatal testing (NIPT) has become an important tool in prenatal screening, yet the question of “when to test for the best outcome” in clinical practice still largely relies on rule-of-thumb time windows, and the impact of individual differences—especially maternal BMI—on testing effectiveness is often underestimated. This study proposes an individualized framework for determining the testing time that proceeds from dependence identification, nonlinear modelling, segmented risk evaluation, and global optimization.” We first use Spearman correlation and distance correlation to characterize the dependence structure between key variables and Y-chromosome concentration, and then adopt a generalized additive model (GAM) to obtain an interpretable nonlinear baseline. Building on this, we derive optimal BMI segments via dynamic programming on model residuals, and construct an integrated risk function that simultaneously covers false negatives, false positives, and test failure, while incorporating a gestational-age penalty and an adjustment factor for “attainment status.” Finally, we perform a global search with simulated annealing over the feasible gestational window to obtain the optimal testing time for each segment. Empirical results show that, under the baseline scenario, the optimal testing times for the six BMI groups cluster around 10.0–11.7 weeks, which are overall substantially earlier than the actual testing weeks (by about 4.4–8.4 weeks). Risk decomposition indicates that the delay penalty dominates, and test-failure risk is higher in high-BMI groups. Under multiple error scenarios of light/moderate/severe perturbations, the optimal time essentially converges to around 10.0 weeks, indicating robust decision-making. For female-fetus samples, the constructed SVM classifier achieves an AUC of 0.9550 without relying on Y-chromosome information. The framework provides a reproducible and practical quantitative basis for individualized NIPT testing-time recommendations and female-fetus abnormality classification.

Keywords

Non-invasive prenatal testing (NIPT); BMI segmentation; generalized additive model (GAM); risk function; simulated annealing; female fetus abnormality classification