This retrospective study was conducted using medical records of women with placenta previa accreta in a single tertiary hospital from January 1, 2015, to December 31, 2020. Placenta previa was diagnosed through routine prenatal ultrasound in the third trimester by experienced ultrasonographers. Ultrasonographic diagnosis of placenta accreta was based on previously reported criteria [13,14,15] when at least one of the following abnormal ultrasound findings were present: loss of retroplacental clear zone, myometrial thinning, multiple placental lacunae, subplacental hypervascularity, turbulent flow in lacunae and bladder wall interruption. Magnetic resonance imaging (MRI) was performed according to the managing obstetrician’s discretion. Characteristics of MRI considered suggestive of placenta accreta included the following [16,17,18]: uterine bulging, intraplacental dark bands on T2-weighted imaging, heterogeneous placental signal, bladder tenting, placental protrusion into the cervix, or more than one of these characteristics. Placenta accreta was confirmed during surgery when there was a lack of spontaneous complete separation of the placenta from its basal plate, requiring manual removal or after direct visualization of placental tissue protruding through the uterine serosa. The degree of placenta accreta was diagnosed by clinical criteria and histologic criteria according to the International Federation of Gynecology and Obstetrics (FIGO) classification from 2018 . The inclusion criteria were women diagnosed with placenta previa and suspicious accreta by ultrasound or MRI before caesarean delivery who were confirmed to have accreta after caesarean delivery. The exclusion criteria were multiple pregnancies and women with a delivery week < 28 weeks.
A total of 156 women participated in this study (Fig. 1). They were divided into a balloon group (n = 68) and a control group (n = 88) based on whether they received AABO. Prophylactic placement of balloon catheters was performed at the surgeon’s discretion. All women who received balloon catheters provided written informed consent. For all women in the control group, written informed consent was signed for caesarean delivery without prophylactic placement of balloon catheters. For women in the balloon group, the balloon catheter was performed before surgery, and the procedure was as follows. The catheter was inserted via the right femoral artery and placed at the abdominal aorta (above the division of the common iliac artery and under the renal ostia from the aorta) under fluoroscopic guidance with the balloon deflated. Then, the patient was transferred to the operating room for caesarean delivery. The balloon was inflated immediately after cord clamping. The balloon was inflated for no longer than 10 min and was then deflated for 1 min if the bleeding was active. If the bleeding became inactive, the balloon was kept deflated until the surgery was completed. After delivering the neonate, gentle controlled cord traction was performed in hopes of detaching the placenta if no placenta percreta was suspicious and no heavy bleeding happened. If removal of the placenta failed and the surgeons considered conservative management reasonable, surgical procedures including manual removal of the placenta, clamping and resecting the residual adherent tissue and the affected uterine wall, focal suturing on the placental detachment surface, and the uterine wall reconstruction were performed. Intraoperative haemostatic approaches such as uterine compression sutures, intrauterine gauze packing, intrauterine balloon tamponade and uterine artery ligation could be performed to control bleeding at the surgeon’s discretion. When all conservative therapies failed, the decision to perform a hysterectomy was made by at least two experienced surgeons. The catheter was removed after the surgery, and uterine artery embolisation was adopted if the patient was haemodynamically stable but exhibited persistent slow bleeding. For women in the control group, caesarean delivery was directly performed, and the perioperative management of the two groups was similar.
Maternal characteristics and pregnancy outcomes were collected. Maternal outcomes included operation time, estimated blood loss (EBL), autologous blood transfusion, homologous transfusion including packed red blood cells (PRBCs) transfusion, fresh frozen plasma (FFP) transfusion and other blood transfusions, intraoperative uterine compression suture, intrauterine gauze packing, intrauterine balloon tamponade, intraoperative bladder injury, hysterectomy, disseminated intravascular coagulation (DIC), haemorrhagic shock, uterine artery embolisation after surgery, relaparotomy, maternal length of antibiotic use, maternal length of hospital stay after surgery, complications of thromboembolism diseases, haematoma and artery rupture. The amount of intraoperative EBL was determined by adding blood volume in the surgical suction bottle (excluding amniotic fluid), that on gauze, and the visually estimated volume on the operating table. The primary outcomes included EBL ≥ 2.0 L, massive transfusion (transfusion of 10 or more units of packed red blood cells) and hysterectomy. Neonatal outcomes included gestational weeks at delivery, birthweight, Apgar score ≤ 7 at 1 min, Apgar score ≤ 7 at 5 min, neonatal intensive care unit (ICU) admission, assisted ventilation, pulmonary surfactant use, the complication of neonatal respiratory distress syndrome (NRDS), wet lung, neonatal asphyxia, neonatal infection, hyperbilirubinemia, antibiotic administration and neonatal length of hospital stay.
Data were analysed by SPSS statistical software version 22.0. Quantitative variables are presented as the mean ± standard deviation (SD), and qualitative variables are presented as frequencies and percentages. Continuous data were compared using Student’s t test or Wilcoxon rank-sum test for nonnormally distributed continuous variables. Categorical data were compared using the x2 test or Fisher’s exact test. A propensity score analysis with 1:1 matching was applied to minimise the indication bias of the two groups. Univariate analyses were adopted to screen characteristics with P < 0.2, and then a multivariable logistic regression model was adopted to acquire independent predictors for the composite primary outcome. P < 0.05 was considered statistically significant.