Development of a spontaneous preterm birth predictive model using a panel of serum protein biomarkers for early pregnant women: A nested case-control study

Abstract

Objective: To develop a model based on first trimester maternal serum LC-MS/MS to predict spontaneous preterm birth (sPTB) < 37weeks. Methods: A cohort of 2,053 women were enrolled in a tertiary maternity hospital in China from July 1, 2018 to January 31, 2019. In total, 110 singleton pregnancies (26 cases of sPTB and 84 controls) at 11-136/7 gestational weeks were used for model development and internal validation. A total of 72 pregnancies (25 cases of sPTB and 47 controls) at 20-32 gestational weeks from an additional cohort of 2,167 women were used to evaluate the scalability of the prediction model. Maternal serum samples were collected at enrollment and analyzed by LC-MS/MS, and candidate proteins were used to develop an optimal predictive model by machine learning algorithms. Results: A novel predictive panel with four proteins, including sFlt-1, MMP-8, ceruloplasmin, and SHBG, which was the most discriminative subset, was developed. The optimal model of logistic regression had an AUC of 0.934, with additional prediction of sPTB in second and third trimester (0.868 AUC). Importantly, higher-risk subjects defined by the prediction generally gave birth earlier than lower-risk subjects. Conclusion: First trimester modeling based on maternal serum LC-MS/MS identifies pregnant women at risk of sPTB, which may provide utility in identifying women at risk at an early stage of pregnancy before clinical presentation to allow for earlier intervention.