基于三维能量多普勒超声联合PLGF的子痫前期风险预测模型的建立

Establishment of preeclampsia risk prediction model based on 3D power Doppler ultrasound combined with PLGF

目的建立基于三维能量多普勒超声联合胎盘生长因子(placental growth factor,PLGF)的子痫前期(pre-eclampsia,PE)风险预测模型,为早期有效预测PE提供参考。方法选择2018年1月至2022年1月在深圳市南山区妇幼保健院收治的163例PE患者为病例组,选择同期本院孕检的健康孕妇163例为对照组进行回顾性分析。收集并比较两组研究对象的一般资料、三维能量多普勒超声检查指标及实验室指标,以LASSO回归筛选变量后行多因素Logistic回归筛选出独立影响因素,根据Logistic回归结果构建列线图模型,并对模型进行验证。结果两组年龄、孕前BMI、孕期增重、高血压病家族史、吸烟史、羊水过少、血管指数(vascularization index,VI)、血流指数(flow index,FI)、血管血流指数(vascularization flow index,VFI)、血红蛋白(hemoglobin,HB)、红细胞压积(hematocrit,HCT)及PLGF差异均有统计学意义(P<0.05)。LASSO筛选出7个潜在的预测因素分别为高血压病家族史、年龄、VI、FI、VFI、HCT及PLGF。多因素二元Logistic逐步回归结果显示:高血压病家族史、年龄、VI、FI、VFI、HCT及PLGF为PE发生的独立影响因素(P<0.05)。给予多因素分析结果以R软件构建列线图模型,ROC法评估列线图模型对PE的预测效能,AUC为0.905,95%CI为0.873-0.938。校准曲线结果显示:预测曲线与标准曲线基本拟合,H-L拟合优度检验结果显示χ^(2)=0.426,P=0.514。结论PE的发生与孕妇高血压病家族史、年龄、三维能量多普勒超声指标等因素有关,采用上述指标构建的列线图模型可用于PE的预测且具有较高的准确度与区分度。

Objective To establish a preeclampsia(PE)risk prediction model based on three-dimensional power Doppler ultrasound combined with placental growth factor(PLGF),to provide a reference for early effective prediction of PE.Methods 163 PE patients in Shenzhen Nanshan Maternal and Child Healthcare Hospital from January 2018 to January 2022 were selected as the case group,and 163 healthy pregnant women in our hospital during the same period were selected as the control group for retrospective analysis.The general data,three-dimensional power Doppler ultrasound examination indexes and laboratory indexes of the two groups were collected and compared.LASSO regression was used to screen variables,and then multivariate Logistic regression was performed to screen out independent influencing factors.Constructed a column chart model based on Logistic regression results and validate the model.ResulstsThere were significant differences in age,pre-pregnancy BMI,weight gain during pregnancy,family history of hypertension,smoking history,oligohydramnios,vascularization index(VI),flow index(FI),vascularization flow index(VFI),hemoglobin(HB),hematocrit(HCT)and PLCF between the two groups(P<0.05).Seven potential predictors were screened by LASSO:family history of hypertension,age,VI,FI,VFI,HCT and PLGF.Multivariate binary Logistic regression showed that family history of hypertension,age,VI,FI,VFI,HB,HCT and PLGF were independent influencing factors of PE(P<0.05).Given the multivariate analysis results,a nomogram model was constructed with R software,and the ROC method was used to evaluate the predictive performance of the nomogram model for PE,the AUC was 0.905,and the 95%CI was 0.873-0.938.The results of the calibration curve showed that the predicted curve and the standard curve were basically fitted,and the results of the H-L goodness-of-fit test showed that the χ^(2)=0.426,P=0.514.Conclusion The occurence of PE is related to factors such as maternal family history of hypertension,age and three-dimensional power Doppler ultrasound indexes.The nomogram model constructed by the above indexes can be used to predict PE with high accuracy and discrimination.