妊娠相关血栓性微血管病预测模型的建立与验证

Establishment and validation of a predictive model for pregnancy-related thrombotic microangiopathy

目的 妊娠相关血栓性微血管病(P-TMA)是围生期的急危重症,起病急骤突然,常危及母婴安全,治疗手段有限。寻求非创伤性的P-TMA预测方法,对于改善母婴预后具有重要的意义。方法 通过回顾性分析建立预测模型,选用临床诊断P-TMA患者61例,同时选用520例健康孕产妇作为对照,比较两组实验室指标的差异,依据logistic回归,建立P-TMA预测模型。通过前瞻性研究对预测模型进行外部验证。入选28周以上孕产妇志愿者895例,用建立的预测模型进行测算,其中预测阳性67例,阴性828例。随访至产后3个月,验证和评估该预测模型的价值。结果 血小板(PLT)、乳酸脱氢酶(LDH)及血管性血友病因子(vWF)为P-TMA病变的有效预测指标(均P<0.05),并据此建立logistic回归方程。该方程对P-TMA病变的预测概率拟合ROC曲线,AUC为0.997(P<0.001)。Hosmer-Lemeshow检验的拟合度良好(P=0.986)。在895例志愿者中进行的前瞻性外部验证,AUC为0.830(P<0.001),预测阳性到临床诊断的时间间隔为(11±6.5)d,提示该模型对P-TMA具有良好的预测功能。结论 基于PLT、LDH及vWF建立的无创性的P-TMA预测模型,临床易开展,可有效对P-TMA患者早期预警,改善母婴预后。

Objective Pregnancy-related thrombotic microangiopathy(P-TMA)is an acute and critical illness in the perinatal peri-od,with abrupt onset,often endangering the safety of mothers and infants,and limited treatment options.Finding a non-invasive P-TMA prediction method is of great significance for improving maternal and infant prognosis.Methods A predictive model was established through retrospective analysis,61 patients with clinical diagnosis of P-TMA were selected,and 520 healthy pregnant women were selected as controls to compare the differences in laboratory indicators between the two groups,and establish a P-TMA prediction model based on logistic regres-sion.The predictive model was externally validated through prospective studies.A total of 895 volunteers with over-28-week preganacy were selected,and the established prediction model was used for calculation,of which 67 were positive and 828 were negative.The value of this predictive model was validated and evaluated with follow-up up to 3 months postpartum.Results Platelet(PLT),lactate dehydrogenase(LDH)and von Willebrand factor(vWF)were effective predictors of P-TMA lesions(all P<0.05),and a logistic regression equation was established accordingly.The equation fit the ROC curve to the predicted probability of P-TMA lesions,and the area under ROC curve(AUC)was 0.997(P<0.001),with good fit of the Hosmer-Lemeshow test(P=0.986).In the prospective external validation in 895 vol-unteers,the AUC was 0.830(P<0.001),and the time interval from positive prediction to clinical diagnosis was(11±6.5)days,sugges-ting that the model has a good predictive performance for P-TMA.Conclusion The non-invasive P-TMA prediction model established on PLT,LDH and vWF is easy to carry out clinically,and can effectively provide early warning for P-TMA patients and improve the prognosis of mother and child.