Predicting the probability of a vaginal birth after previous cesarean section: validation of two prediction models in a western European cohort

ObjectiveIn the western European population, there is no model available for prediction of the probability of achieving vaginal birth after a previous cesarean section (VBAC). Our objective was to validate, in a Dutch cohort, the entry-to-care and close-to-delivery prediction models by Grobman et al. [1,2].Study DesignWe composed a retrospective cohort from patient records in 17 representative hospitals in the Netherlands. We included women with a history of one previous low-transverse cesarean section (CS) and viable singleton cephalic pregnancy, who attempted trial of labor (TOL). We evaluated the performance of both models in the prediction of VBAC, which was assessed by the area under the receiver operating characteristic curve (AUC) and calibration per risk decile.ResultsOf the 757 women eligible for TOL, 248 had a repeat CS and 509 had a TOL, the observed VBAC-rate was 72.2%. The AUC's of the two models were 0.66 (95% CI 0.60-0.72) and 0.68 (95% CI 0.62-0.74) respectively (p-value for difference p=0.058). These values were slightly lower than those obtained in the original models development. Despite the models relatively mediocre performance in classifying women into a dichotomous outcome (i.e. VBAC vs. repeat CS), the calibration with regard to clinically-relevant categories was good. The observed rates were highly correlated with the predicted rates, although there was a general tendency toward underestimation of the actual VBAC rates which was more pronounced in the entry-to care model. The model demonstrated the potential to re-classify a notable portion of patients away from the population mean.ConclusionBoth prediction models are useful for discriminating patients with regard to whether they have a high or low chance of achieving VBAC. References: [1]Grobman WA et al. Development of a nomogram for prediction of vaginal birth after cesarean delivery. Obstet Gynecol. 2007 [2]Grobman WA et al. Does information available at admission for delivery improve prediction of vaginal birth after cesarean? Am J Perinatol. 2009 ObjectiveIn the western European population, there is no model available for prediction of the probability of achieving vaginal birth after a previous cesarean section (VBAC). Our objective was to validate, in a Dutch cohort, the entry-to-care and close-to-delivery prediction models by Grobman et al. [1,2]. In the western European population, there is no model available for prediction of the probability of achieving vaginal birth after a previous cesarean section (VBAC). Our objective was to validate, in a Dutch cohort, the entry-to-care and close-to-delivery prediction models by Grobman et al. [1,2]. Study DesignWe composed a retrospective cohort from patient records in 17 representative hospitals in the Netherlands. We included women with a history of one previous low-transverse cesarean section (CS) and viable singleton cephalic pregnancy, who attempted trial of labor (TOL). We evaluated the performance of both models in the prediction of VBAC, which was assessed by the area under the receiver operating characteristic curve (AUC) and calibration per risk decile. We composed a retrospective cohort from patient records in 17 representative hospitals in the Netherlands. We included women with a history of one previous low-transverse cesarean section (CS) and viable singleton cephalic pregnancy, who attempted trial of labor (TOL). We evaluated the performance of both models in the prediction of VBAC, which was assessed by the area under the receiver operating characteristic curve (AUC) and calibration per risk decile. ResultsOf the 757 women eligible for TOL, 248 had a repeat CS and 509 had a TOL, the observed VBAC-rate was 72.2%. The AUC's of the two models were 0.66 (95% CI 0.60-0.72) and 0.68 (95% CI 0.62-0.74) respectively (p-value for difference p=0.058). These values were slightly lower than those obtained in the original models development. Despite the models relatively mediocre performance in classifying women into a dichotomous outcome (i.e. VBAC vs. repeat CS), the calibration with regard to clinically-relevant categories was good. The observed rates were highly correlated with the predicted rates, although there was a general tendency toward underestimation of the actual VBAC rates which was more pronounced in the entry-to care model. The model demonstrated the potential to re-classify a notable portion of patients away from the population mean. Of the 757 women eligible for TOL, 248 had a repeat CS and 509 had a TOL, the observed VBAC-rate was 72.2%. The AUC's of the two models were 0.66 (95% CI 0.60-0.72) and 0.68 (95% CI 0.62-0.74) respectively (p-value for difference p=0.058). These values were slightly lower than those obtained in the original models development. Despite the models relatively mediocre performance in classifying women into a dichotomous outcome (i.e. VBAC vs. repeat CS), the calibration with regard to clinically-relevant categories was good. The observed rates were highly correlated with the predicted rates, although there was a general tendency toward underestimation of the actual VBAC rates which was more pronounced in the entry-to care model. The model demonstrated the potential to re-classify a notable portion of patients away from the population mean. ConclusionBoth prediction models are useful for discriminating patients with regard to whether they have a high or low chance of achieving VBAC. References: [1]Grobman WA et al. Development of a nomogram for prediction of vaginal birth after cesarean delivery. Obstet Gynecol. 2007 [2]Grobman WA et al. Does information available at admission for delivery improve prediction of vaginal birth after cesarean? Am J Perinatol. 2009 Both prediction models are useful for discriminating patients with regard to whether they have a high or low chance of achieving VBAC. References: [1]Grobman WA et al. Development of a nomogram for prediction of vaginal birth after cesarean delivery. Obstet Gynecol. 2007 [2]Grobman WA et al. Does information available at admission for delivery improve prediction of vaginal birth after cesarean? Am J Perinatol. 2009