Prediction of risk for early or very early preterm births using high-resolution urinary metabolomic profiling

This study aimed to predict early (<35 weeks) and very early (≤26 weeks) PTB using urinary metabolomics from samples collected between weeks 8 to 24 of pregnancy. Models were developed using samples from Stanford University (30 subjects) and validated with samples from the University of Alabama (24 subjects). Metabolomic profiles were analyzed using mass spectrometry and predictive models built using XGBoost. Among 7,913 features, 12 metabolites were identified and used in models achieving AUROCs of 0.995 and 0.964 (early PTB), and 0.950 and 0.830 (very early PTB) for development and validation cohorts, respectively, highlighting their potential for early risk assessment.