Odibo Anthony O, Cahill Alison G, Stamilio David M, Stevens Erika J, Peipert Jeffrey F, Macones George A
Division of Maternal Fetal Medicine and Ultrasound, Department of Obstetrics and Gynecology, Washington University, St. Louis, Missouri 63110, USA.
Am J Perinatol. 2007 May;24(5):299-305. doi: 10.1055/s-2007-981430. Epub 2007 May 18.
The purpose of this study was to determine if placental abruption or previa in women with a history of a prior cesarean delivery (CD) can be predicted. A retrospective cohort study of pregnant women with previous CD was conducted in 17 centers between 1996 and 2000. Women developing placenta previa or abruption in the subsequent pregnancy were compared with those without these complications. Bivariate and multivariable techniques were used to develop predictive models for placenta previa or abruption. The area under the receiver-operator characteristic curves, sensitivity, specificity, and accuracy of the models were compared. Among 25,076 women with prior CD, there were 361 (15 per 1000 births) with placenta previa and 309 (13 per 1000 births) with abruption. The significant risk factors for these complications include advanced maternal age, Asian race, increased parity, illicit drug use, history of spontaneous abortion, and three or more prior cesarean deliveries. Prediction models for abruption and previa had poor sensitivity (12% and 13% for abruption and previa, respectively). In women with at least one prior cesarean delivery, the risk factors for placental previa and abruption can be identified. However, prediction models combining these risk factors were too inefficient to be useful.
本研究的目的是确定有剖宫产史的女性是否可预测胎盘早剥或前置胎盘。1996年至2000年间，在17个中心对有既往剖宫产史的孕妇进行了一项回顾性队列研究。将后续妊娠中发生前置胎盘或胎盘早剥的女性与未发生这些并发症的女性进行比较。采用双变量和多变量技术建立前置胎盘或胎盘早剥的预测模型。比较了模型的受试者工作特征曲线下面积、敏感性、特异性和准确性。在25076例有既往剖宫产史的女性中，有361例（每1000例分娩中有15例）发生前置胎盘，309例（每1000例分娩中有13例）发生胎盘早剥。这些并发症的显著危险因素包括产妇年龄较大、亚洲种族、产次增加、使用非法药物、自然流产史以及三次或更多次既往剖宫产史。胎盘早剥和前置胎盘的预测模型敏感性较差（胎盘早剥和前置胎盘的敏感性分别为12%和13%）。在至少有一次既往剖宫产史的女性中，可以识别出前置胎盘和胎盘早剥的危险因素。然而，结合这些危险因素的预测模型效率太低，无法发挥作用。
