Introduction

Preeclampsia (PE) is acknowledged as a significant contributor to both maternal and fetal mortality and is linked to a considerably heightened risk of long-term cardiovascular and chronic conditions after pregnancy, such as chronic hypertension, heart disease, stroke, metabolic syndrome, cognitive decline, and end-stage renal failure.1, 2, 3 Despite extensive research, the pathogenesis of PE remains unclear, and no effective treatment exists beyond delivery. Studies have shown that administering low-dose aspirin (50–150 mg/day) before 16 weeks of gestation can reduce the incidence of preterm PE by nearly 60%.4,5 Therefore, early identification of high-risk women is crucial for timely intervention and improved pregnancy outcomes.

The Fetal Medicine Foundation (FMF) models, which combine maternal factors, mean arterial pressure (MAP), uterine artery pulsatility index (UtA-PI), and placental growth factor (PLGF), are widely validated but have limited practicality due to complex assessments and specialized testing. Moreover, the FMF model detects approximately 75% of preterm PE cases (delivery before 37 weeks) and 40%–45% of term PE cases at a false-positive rate of 10%, leaving room for improvement.6 Routine laboratory markers assessing liver, kidney, lipid, glucose, and trace element profiles are standard in prenatal screening. PE is often accompanied by hepatic impairment, renal dysfunction, and metabolic disturbances, reflected in abnormalities such as elevated bilirubin, alanine aminotransferase (ALT), aspartate aminotransferase (AST), and uric acid (UA) levels. Yet, whether these markers exhibit predictive variability before PE onset remains understudied. Moreover, insulin resistance has also been identified as a key contributor to PE development.7 The triglyceride-glucose (TyG) index, derived from routine laboratory parameters, is a validated indicator of insulin resistance.8 Although well recognized in glucose metabolism disorders, its predictive value for early PE remains unclear.

To address these gaps, this study investigates whether first-trimester routine laboratory markers can predict PE. Initial analyses focus on identifying laboratory markers associated with PE and determining whether these relationships are linear or nonlinear. Where nonlinear associations are observed, threshold effects are further examined. Leveraging the power of machine learning capable of uncovering complex interactions between variables that traditional linear methods might miss, this study explores the predictive performance of multiple–indicator combinations within laboratory data. Additionally, it evaluates the integration of these findings with existing prediction models to assess potential improvements in predictive efficacy. This research aims to identify novel predictive markers for PE, offering a more accessible and accurate tool for early prevention.