Abstract
OBJECTIVE: This study aims to develop a model to predict preterm birth in patients with velamentous cord insertion (VCI) during the second trimester. METHODS: This retrospective cohort study enrolled 240 singleton pregnancies (135 VCI cases and 105 controls) from January 2023 to December 2024. Four prediction models were constructed using univariate analysis and LASSO regression: Model 1 (midtrimester clinical parameters), Model 2 (late-pregnancy ultrasound parameters), Model 3 (combined mid- and late-pregnancy parameters), and Model 4 (midtrimester growth deviation indices, including BPD(FL) and Z(doppler)). Model performance was assessed using the AUC, sensitivity, specificity, and calibration. A three-tier risk stratification system was established based on predicted probabilities. RESULTS: The preterm birth rate was significantly higher in the VCI group than in the control group (17.8% vs. 5.7%, P < 0.001). Model 4, relying exclusively on midtrimester data (21 ± 2 weeks), achieved an AUC of 0.801 (bootstrap AUC 0.809, 95% CI: 0.710-0.898), with 70.8% sensitivity and 79.3% specificity. While its discriminative performance was lower than that of late-pregnancy-based models (Model 2 AUC 0.972), Model 4 provided approximately 16 weeks of advance warning. Risk stratification identified low-risk (6.5%), moderate-risk (15.6%), and high-risk (46.4%) groups, each demonstrating distinct preterm birth rates. CONCLUSION: The midtrimester ultrasound deviation model enables early identification of high-risk VCI pregnancies, facilitating individualized delivery timing decisions while minimizing unnecessary interventions.