Risk Prediction Models Of Adverse Events In Women With Scarred Uterus And Placenta Previa And Increta Or Percreta

Background:With the increasing of cesarean delivery rates and the implementation of a two-child policy,placenta accreta spectrum(PAS)has been rapidly increased in China,and the following maternal adverse events have raised extensive concern.Placenta increta and percreta lack recognized pre-pregnancy system assessment to predict the maternal outcomes worldwide.So we evaluated the clinical characteristics and perioperative events to find independent predictors which are significantly associated with maternal adverse outcomes,and develop prenatal risk assessment models to predict maternal adverse events in scarred uterus and placenta previa and increta.Methods:This was a retrospective study of all patients with scarred uterus and placenta previa and increta who terminated pregnancy by cesarean section at the Department of Obstetrics,Qilu Hospital of Shandong University between January 2016 and June 2019.All were divided into 2 groups,the development group and the validation group.Clinical characteristics of included maternal and neonatal outcomes were analyzed.Univariate and multivariate logistic regression analysis were used to identify characteristics that were associated with intraoperative massive blood loss(?2500 ml),and massive transfusion of packed red blood cells(?10 U).The corresponding nomogram prediction models were drawn by R software according to the regression coefficients.The receiver operating characteristic curves and calibration curve were applied to assess the discrimination and calibration of the multivariable logistic regression models,respectively.Decision curve analysis was applied to assess the performance of the predictive models.Results:Two hundred and sixty patients with scarred uterus and placenta previa and increta were finally included in the study.Among them,there were 240(92.3%)patients with placenta increta and 20(7.7%)patients with placenta percreta.192(73.8%)had no more than one previous cesarean section,and 68(26.2%)had more than one previous cesarean section.Seventy-one(27.3%)patients lost massive blood volume(?2500 ml),and of whom 8(3.1%)patients performed hysterectomy;55(21.2%)patients transfused massive volume of packed red blood cells(? 10 U).Based on the preoperative feature group and preoperative feature and surgical clinical feature group,4 models were constructed to predict 2 adverse events.Model 1 to predict massive blood loss has 6 independent predictors,including vascular lacunae within the placenta(p=0.005),hypervascularity of uterine-placental margin(p=0.002),hypervascularity of the uterine serosa-bladder wall interface(p=0.020),hypervascularity of cervix(p=0.016),emergency cesarean section(p<0.001),and no preoperative balloon placement of abdominal aorta(p<0.001).Model 2 to predict massive blood loss has 4 independent predictors,including vascular lacunae within the placenta(p=0.006),hypervascularity of uterine-placental margin(p=0.016),hypervascularity of cervix(p=0.036),emergency cesarean section(p<0.001).Model 3 to predict massive blood transfusion has 6 independent predictors,including vascular lacunae within the placenta(p=0.001),hypervascularity of uterine-placental margin(p=0.003),hypervascularity of the uterine serosa-bladder wall interface(p=0.004),hypervascularity of cervix(p=0.027),emergency cesarean section(p=0.001),and no preoperative balloon placement of abdominal aorta(p<0.001).Model 4 to predict massive blood transfusion has 4 independent predictors,including vascular lacunae within the placenta(p=0.004),hypervascularity of uterine-placental margin(p=0.017),hypervascularity of the uterine serosa-bladder wall interface(p=0.023),and no preoperative balloon placement of abdominal aorta(p=0.001).In the development group and the validation group,the area under the receiver operating characteristic curve of model 1 has a better discrimination capacity compared with that of model 2(p<0.001;p=0.007),model 3 has a better discrimination capacity compared with model 4(p=0.001;p=0.015).The calibation plots demonstrated a good agreement between the predicted risks and observed incidence of the above 4 models.The net benefit of the model 1 was better than model 2 in DCA,the net benefit of the model 3 was better than model 4 in DCA.Conclusions:Four prenatal predict assessment models were built successfully,and the model 1 and model 3 have excellent discrimination,calibration,and performance of prediction for predicting massive blood loss and transfusion,respectively.Utilizing the prenatal evaluation models,obstetricians could evaluate placenta increta and percreta patients in a more objective way,and expected to provide a powerful tool for the clinical treatment of pregnant women with increta and percreta.