Prediction Of Preeclampsia By Blood Pressure Trajectories From Latent Class Growth Modeling And Placenta-derived Biomarkers

Objectives:Hypertensive disorders in pregnancy?HDP?,with a prevalence of 5%to 10%of pregnant women,are one of the most important causes for maternal and infant death globally.Moreover,recent evidence showed that the occurrence of HDP is associated with an increased risk of remote cardiometabolic disorders both for the mothers and their babies.HDP is a clinical syndrome characterized by elevated blood pressure?BP?,which can be further classified as chronic hypertension?pre-existing hypertension,defined as BP level?140/90 mmHg before 20 weeks of gestation or before pregnancy?,gestational hypertension?defined as BP level?140/90 mmHg after20 weeks of gestation?,preeclampsia?hypertension plus proteinuria and/or symptoms of organ dysfunction?and eclampsia,as well as chronic hypertension superimposed preeclampsia/eclampsia.Among them,preeclampsia accounts for⁵0%of total HDP cases,which is thought to be a placenta-originated syndrome with unknown causes.Once diagnosed,the only curable approach for preeclampsia is the delivery of placenta.Great efforts have been focused on the development and validation of preeclampsia prediction models,based on demographic information,clinical phenotypes and circulating biomarkers.However,despite limited clinical efficacy,these models are mainly derived from populations from western countries,and may be not applicable in Chinese population.Moreover,numerous placenta-derived biomarkers has been proposed for preeclampsia prediction,their accuracies are not clinical acceptable.By applying latent class growth modeling?LCGM?,we previously identified distinctive BP trajectories associated with excessive BP elevation before term and an increased risk of postpartum metabolic syndrome in normotensive pregnant women,using BP information obtained by routine antepartum examinations.Additionally,recent studies have identified a placenta-derived biomarker,Elebela,in the pathogenesis of preeclampsia in animal models.The predictive value of Elebela for preeclampsia in large birth cohort,has not been validated yet.Therefore,the purpose of the present work lies in two parts:1)in a large cohort enrolling pregnant women at gestational age of¹2 weeks in Tianjin,by using LCGM to model longitudinal obtained BP data from community hospitals,to explore potential BP trajectories during pregnancy and their associations with the risk of future preeclampsia;2)in a nested case-control design,by incorporating serum Elebela level before the occurrence of preeclampsia,as well as demographic and clinical information,to develop prediction model for preeclampsia,and compared its efficacy with other models using established biomarkers.Materials and Methods:?1?We launched a large prospective cohort study from November 1,2016,enrolling pregnant women of¹2 weeks of gestation in 19 community hospitals in Tianjin.Up to May 30,2018),a total of 5809 cases with known pregnant outcomes were documented.After excluding those with 249 cases?42 cases with pregnancy loss before 20 weeks;22 cases with voluntarily withdrew from the study after enrollment;25 cases with no biological sample in early pregnancy;23 cases contact during follow up;41 cases with chronic hypertension;96 cases with twin pregnancies?,5560 cases with singleton pregnancy were included for analysis were used for analysis.Among them,106 cases were diagnosed gestational hypertension,45 cases with preeclampsia with mild features,83 cases with preeclampsia with severe features.?2?In Tianjin maternal and child healthcare system,from 28 weeks of gestation,antepartum documentation of pregnant women will be transferred from community hospital to upper level hospital till delivery.We therefore utilized information obtained during 5 antepartum examinations before week 28,i.e.,week 12,week 16,week 20,week 24 and week 28.?3?We used LCGM to model systolic?SBP?and diastolic BP?DBP?trajectories measured longitudinally.For the association study,the predictors were SBP and DBP trajectory membership?built separately?,the outcomes was the occurrence of preeclampsia that diagnosed in upper level hospital after 28 weeks of gestation.We controlled the influence of other variables that were significantly different at 12weeks of gestation,i.e.,baseline body mass index?BMI?,being primipara or not,white blood cell counts,hemoglobin level,platelet counts and alanine aminotransferase level.?4?For the prediction model building study,we measured serum levels of Elabela from samples obtained at 12 weeks of gestation in 128 preeclampsia patients,as well as in 88 normotensive subjects without antepartum maternal and neonatal complications.The outcome was the occurrence of preeclampsia.The predictors were serum Elabela level,BP level at 12 weeks,baseline BMI and being primipara or not.?5?Prediction models were also built based on other established biomarkers?soluble fms-like tyrosin kinase-1,sFlt-1;and placenta growth factor,PlGF?,by incorporating BP level at 12 weeks,baseline BMI,being primipara as well.The discrimination and calibration of fitted models were evaluated by using C-statistic in Logistic regression models and Hosmer-Lemeshow test,respectively.Internal validation of the prediction models was performed by bootstrapping,using 1000 random resamples from the entire data set.Results:?1?For SBP trajectory identification,by using LCGM,we identified 4 distinctive SBP trajectories from 12 weeks to 28 weeks of gestation.Depending on the SBP level from upper to lower levels,the four trajectories had specifications of linear?designated as SBP_traj₄,accounts for 7.6%of the total population?,cubic?SBP_traj₃:35.1%?,quadratic?SBP_traj₂:48.4%?,and intercept only?SBP_traj₁:8.8%?,respectively.?2?Logistic regression showed that the higher the trajectory located,the higher risk for preeclampsia:using the combination of SBP_traj₁ and SBP_traj₂ as the reference group?due to no case of preeclampsia in SBP_traj₁ was reported?,the odds ratio?OR?for SBP_traj₄ was 6.678?95%confidence interval?CI?:4.106 to10.861,P<0.001?,and the OR for SBP_traj₃ was 2.368?95%CI:1.583 to 3.542,P<0.001?.After controlling for potential confounders?baseline BMI,being primipara or not,white blood cell counts,hemoglobin level,platelet counts and alanine aminotransferase level?,the OR for SBP_traj₄ was 4.023?95%CI:2.368 to 6.835,P<0.001?,and the OR for SBP_traj₃ was 1.854?95%CI:1.223 to 2.811,P=0.004?.?3?For DBP trajectory modeling,we identified 4 distinctive trajectories from 12weeks to 28 weeks of gestation.Depending on the DBP level at 12 weeks of gestation from the highest to the lowest,the four trajectories had specifications of intercept only?designated as DBP_traj₄,accounts for 13.1%of the total population?,quadratic?DBP_traj₃:24.7%?,quadratic?DBP_traj₂:16.4%?,and linear?DBP_traj₁:45.8%?,respectively.Notably,the DBP_traj₂ had a crossover with DBP_traj₃ at around 22 weeks of gestation,and exceeded DBP_traj₃ thereafter.?4?Logistic regression showed:using the DBP_traj₁ as the reference group,the OR for DBP_traj₄ was 4.100?95%CI:2.571 to 6.538,P<0.001?,and the OR for DBP_traj₃ was 1.582?95%CI:0.970 to 2.578,P=0.066?,and 2.632?95%CI:1.570to 4.414,P<0.001?for DBP_traj₂.After controlling for potential confounders?baseline BMI,being primipara or not,white blood cell counts,hemoglobin level,platelet counts and alanine aminotransferase level?,the OR for DBP_traj₄ was 2.527?95%CI:1.534 to 4.162,P<0.001?,and the OR for DBP_traj₃ was 1.297?95%CI:0.790 to 2.128,P=0.303?,and 2.238?95%CI:1.328 to 3.772,P=0.002?for DBP_traj₂.Therefore,BP trajectories from 12 weeks to 28 weeks identified using LCGM are novel risk factors that independently associated with the occurrence of preeclampsia.?5?For preeclampsia prediction model building study,the candidate variables included three parts:1)biomarker component?Elabela,sFlt-1,PlGF,and the ratio of sFlt-1/PlGF measured at 12 weeks of gestation?;2)BP level component?SBP at 12weeks of gestation;DBP level was not significantly different at this time point?;and 3)demographic information?significant different between preeclampsia and controls:baseline BMI and being primipara or not?.The basic model only incorporated SBP and demographic information components.?6?Compared with basic model,the incorporation of Elabela significant improved model discrimination(C-statistic:model_Elabelalabela 0.791 versus model_basic0.696,P=0.0028).Moreover,compared with models incorporating either sFlt-1 or sFlt-1/PlGF,the discrimination capacity of model with Elabela was also significantly improved(C-statistic:model_Elabelalabela 0.791 versus model_sFlt-10.706,and versus model_sFlt-1/PlGFFlt-1/PlGF 0.709;P=0.0108 and 0.0133,respectively).?7?By using a cutoff value of 2.13?ng/mL?for Elabela,a prediction model was built with clinically acceptable discrimination?C-statistic 0.747,P<0.001?and well-calibrated?Hosmer-Lemeshow test chi-squared=8.51,P=0.385?.The predictive value was further internally validated by bootstrapping method.Conclusion:In a large prospective cohort,we have the following main conclusions.?1?By applying LCGM,we for the first time identified distinctive BP trajectories from gestational week 12 to 28,which can independently predict the development of preeclampsia after 28 weeks of gestation.?2?We for the first time confirmed the predictive value of serum Elabela measured at early pregnancy for preeclampsia occurrence after 28 weeks of gestation.These novel risk factors for preeclampsia need to be validated externally in other studies.