Echocardiography and Biomarker-based Regression Models for the Detection of Gestational Hypertension in Pregnant Women

Abstract

Hypertensive disorders of pregnancy are characterized by both higher cardiovascular risk for women and more pronounced structural and functional changes of the heart. Certain biomarkers and echocardiographic parameters could be utilized in regression models for more accurate detection of the presence of those hypertensive disorders. The aim of the study was to compare the differentiating abilities of several regression models for statistical certainty in establishing the diagnosis of gestational hypertension in pregnant women using echocardiographic parameters and biomarkers. A prospective, single-centre, clinical-epidemiological study was conducted with the participation of 36 women with gestational hypertension and 50 maternal and gestational age-matched healthy pregnant controls. Certain echocardiographic parameters and serum biomarkers – placental growth factor (PlGF), galectin-3, high-sensitivity C-reactive protein and Interleukin-6, were analyzed for the women using ROC curve analysis and binary logistic regression. We constructed three regression models, allowing very good differentiation between women with gestational hypertension and those with normotensive pregnancy. The model that included PlGF, left ventricular and right ventricular global longitudinal strain (LV GLS, RV GLS) had AUC 0.81, sensitivity and specificity of 86% and accuracy of 86%. The LV GLS and PlGF model had AUC of 0.90 with sensitivity 83%, specificity 88%, and accuracy 86%; and the LV GLS and RV GLS had AUC of 0.83, sensitivity 67%, specificity 92%, and accuracy 81%. All three combined models showed better accuracy compared to PlGF, LV GLS, RV GLS or any other echocardiographic parameter or biomarker alone, which could be promising for their implementation in clinical practice.