First-trimester preeclampsia prediction model via integrative machine learning of maternal risk profiles and laboratory markers.

Preeclampsia (PE) affects 2%-8% of pregnancies worldwide and remains a major cause of maternal and perinatal complications. Early risk identification is essential, yet current models, such as those from the Fetal Medicine Foundation (FMF), show limited sensitivity, especially for term PE. We investigated whether first-trimester routine laboratory markers could enhance PE prediction in a retrospective cohort of 12,715 pregnancies, including 556 (4.4%) PE cases. Elevated gamma-glutamyltransferase (GGT), C-reactive protein (CRP), triglyceride-glucose index multiple of the median (TyG MoM), and uric acid-to-albumin ratio multiple of the median (UAR MoM) were independently associated with a higher risk of PE, whereas magnesium, iron, and high-density lipoprotein cholesterol were inversely associated with the condition. Several markers also exhibited nonlinear and threshold effects. These features were incorporated into machine learning frameworks, with distinct models for preterm and term PE. For preterm PE, the CatBoost model achieved an AUC of 0.954 and detected 92.5% of cases at a 14% false-positive rate when integrated with the FMF model, compared with 65% detection with FMF alone. For term PE, logistic regression yielded an AUC of 0.913 and 81.1% detection, substantially outperforming FMF sensitivity (47.7%). Parallel cell-free DNA transcriptomic profiling revealed early metabolic and inflammatory dysregulation, and GEO dataset analyses provided external validation, supporting laboratory findings. These results indicate that the incorporation of accessible laboratory markers into existing clinical frameworks, combined with machine learning, provides a cost-effective and scalable strategy for enhanced PE risk stratification and precision management.