Outcomes of prospectively-collected consecutive cases of antenatal-suspected placenta accreta

Background: Accurate diagnosis of placenta accreta is tentative before surgery. This study developed a predictive score for antenatal diagnosis of placenta accreta through mathematical modeling using clinical signs. Methods: Antenatal cases of suspected placenta accreta were collected prospectively in a single-site tertiary delivery center. Women with clinical signs of placenta accreta (placenta previa, number of previous cesarean deliveries and/or ultrasound suspicion of placenta accreta) were included. The diagnosis of accreta was confirmed surgically. The primary endpoint was the proportion of surgically-diagnosed placenta accreta among all suspected cases. Logistic regression modeling was performed to assess preoperative risk factors for placenta accreta. The risk score was tested on a receiver operator characteristic curve to identify subjects with placenta accreta and the optimum cut-point was chosen. Results: Over nine years, 92 suspected accreta cases were identified from 46623 deliveries (0.2%). The diagnosis was confirmed at surgery in 52/92 cases (56%) and there were no maternal deaths. Blood transfusion requirements were greater in patients with placenta accreta versus patients without placenta accreta (median 7 [range 0-25, interquartile range 3-10] versus 0 [0-6, 0-2] units of blood, P <0.0001). Area under the curve of the receiver operator characteristic curve was 0.846, with contribution from three variables (placenta previa, number of previous cesarean deliveries and ultrasound suspicion), each with a P value <0.05. From the ROC curve a cut-point with 100% sensitivity and specificity 25% (95% CI 12.69%-41.20%) was achieved, compared with 86.6% sensitivity (95% Cl 74.21%-94.41%) and 60.0% specificity (95% CI 43.33%-75.14%) using ultrasound alone. Conclusions: Combining diagnostic features associated with placenta accreta through mathematical modeling has better positive predictive value than ultrasound alone. (C) 2013 Elsevier Ltd. All rights reserved.