328: Prediction of success in external cephalic version for breech presentation at term

Abstract

We aimed at designing a clinically based predictive model for the likelihood of successful external cephalic versions (ECVs) near term. This single-center retrospective study was conducted in a single center between 2016 and 2018 and included all candidates for ECV between 35 and 41 weeks' gestation. The versions were performed by a single experienced obstetrician. Variables that may potentially affect the success rate of external version were collected for all participants and included: BMI, AFI, gestational age, parity, location of placenta, fetal trunk posture, period of time in breech presentation prior to the procedure and the size of the amniotic fluid preceding the fetal presenting part (fore-bag). Patients' characteristics were compared between successful and unsuccessful versions based on multiple logistic regression models in order to generate a discriminative model depicted by a conditional inference classification tree (ctree). Overall 250 pregnant women were eligible and opted for a trial of ECV with a success rate of 64.8%. Variables such as location of the placenta, period of time at breech presentation and previous uterine scar had no statistically significant impact on the success rate of ECV while BMI, size of fore-bag and parity were independent determinants of the final presentation. Prediction model for the outcome of ECV based on the most discriminatory factors is suggested and shown as a classification tree. According to our model, patients with a forebag size of 0 had a predicted probability of 10% for successful version, patients with a forebag size >0 and a BMI >29 had a probability of 0%, patients with a forebag size >1 and BMI <=29 had a probability of 96.3% and patients with forebag size =1 and BMI <=29 and number of previous births>0 had a probability of 81.25%. The model accuracy was 91.9%±5.4% which is significantly better than the no information rate of 64.8% (P<0.001). Sensitivity for predicting successful versions was 0.975 and specificity was 0.818. A prediction model composed of three easily measurable variables enables accurate outcome prediction of ECV at term. Fore-bag was identified as the most important discriminator. Our model holds in internal validation and it can be used to support patient counseling and decision making for ECV.