Abstract
BACKGROUND: This study assessed serum GP73, Nrf2, and CHE levels in pregnancy-associated liver injury and explored their nonlinear risk relationships using restricted cubic spline (RCS) models to establish diagnostic thresholds for early screening. METHODS: Conducted from December 2022 to December 2024, 105 pregnant women with liver injury (research group) and 187 healthy controls were retrospectively analyzed. Multivariate logistic regression, ROC curves, RCS models, and decision curve analysis (DCA) evaluated associations, predictive efficacy, dose-response relationships, and clinical utility of these biomarkers. RESULTS: Elevated GP73 and Nrf2, multiple pregnancies, antipyretic/anti-infective medication use, and malnutrition were risk factors (p < 0.05), while high CHE was protective (p < 0.05). ROC analysis showed AUCs of 0.801 (GP73), 0.785 (Nrf2), and 0.726 (CHE) combined AUC reached 0.951. RCS revealed nonlinear dose-response relationships (p < 0.001), and DCA confirmed clinical significance (net benefit > 0) for individual and combined biomarkers, with the panel showing the highest utility. CONCLUSION: GP73, Nrf2, and CHE demonstrated nonlinear associations with pregnancy-related liver injury, particularly GP73. Combined measurement enhances clinical value for early prediction.