A prediction tool for mode of delivery in twin pregnancies-a secondary analysis of the Twin Birth Study

BACKGROUND: One of the controversies regarding the management of twin gestations relates to the mode of delivery. Currently, counseling regarding the mode of delivery and the chance of successful vaginal twin delivery is based on the average risk for intrapartum cesarean delivery in the general population of twin pregnancies. Decision support tools that provide an individualized risk for intrapartum cesarean delivery based on the unique characteristics of each patient can improve counseling and decision-making regarding the choice of mode of delivery in twin pregnancies. OBJECTIVE: This study aimed to develop and validate a prediction model to determine the risk for intrapartum cesarean delivery in twin pregnancies. STUDY DESIGN: In this secondary analysis of the Twin Birth Study, a multicenter randomized controlled trial, we considered the subgroup of individuals who underwent a trial of vaginal delivery. Candidate predictors included maternal age, parity, previous cesarean delivery, conception method, chorionicity, diabetes and hypertension in pregnancy, gestational age at birth, the onset of labor, presentation of the second twin, sonographic fetal weight estimation, and fetal sex. The co-primary outcomes were overall intrapartum cesarean delivery and cesarean delivery of the second twin. Multivariable logistic regression models were used to estimate the probability of the study outcomes. Model performance was evaluated using measures of discrimination (the area under the receiver operating characteristic curve), calibration, and predictive accuracy. Internal validation was performed using the bootstrap resampling technique. RESULTS: A total of 1221 individuals met the study criteria. The rate of overall intrapartum cesarean delivery and cesarean delivery for the second twin was 25.4% and 5.7%, respectively. The most contributory predictor variables were nulliparity, term birth (>= 37 weeks), a non- cephalic presentation of the second twin, previous cesarean delivery, and labor induction. The models for overall intrapartum cesarean delivery and cesarean delivery of the second twin had good overall discriminatory accuracy (area under the receiver operating characteristic curve, 0.720; 95% confidence interval, 0.688-0.752 and 0.736; 95% confidence interval, 0.669-0.803, respectively) and calibration (as illustrated by the calibration plot and Brier scores of 0.168; 95% confidence interval, 0.156-0.180 and 0.051; 95% confidence interval, 0.040-0.061, respectively). The models achieved good specificity (66.7% and 81.6%, respectively), high negative predictive value (86.0% and 96.9%, respectively), and moderate sensitivity (68.1% and 57.1%, respectively). CONCLUSION: The prediction models developed in this study may assist care providers in counseling individuals regarding the optimal timing and mode of delivery in twin pregnancies by providing individualized estimates of the risk for intrapartum cesarean delivery.