Developing a model of placental lesions prediction in varicose veins patients

Abstract

The objective of the study was to identify predictive biomarkers and generate the model to predict placental lesions in women with varicose veins. We collected serial serum specimens from 128 women with varicose veins between 22 and 24 weeks’ gestation. The investigation includes ultrasound findings, blood analysis of endothelin-1, vascular endothelial growth factor (VEGF), CRP, coagulation factors as well as BMI. We used machine learning algorithm and multivariable logistic regression with Lasso method to predict placental lesions among the pregnant patients with varicose veins.A total of 47 (36.7 %) women with varicose veins subsequently developed placental insufficiency. Mean serum VEGF were higher in women who developed placental insufficiency – 29 (27–31) pg/ml, as compared with women without varicose disease – 24 (22–25) pg/ml, p < 0.001. The performance of the model trained with all the most valuable tests (VEGF, endothelin, CRP, D-dimers, fibrinogen, CEAP class) is admissible (AUC 0.94; CI 0.842–0.956; p < 0.001).We identified novel combination of clinical and laboratory predictive markers that provide pathophysiological insights and could help future improvements of diagnosis and treatment of placental lesions in women with varicose veins.