Examining the validity of a predictive model for vaginal birth after cesarean

Background The Maternal-Fetal Medicine Units (MFMU) vaginal birth after cesarean (VBAC) calculator, while accurate for candidates with high predicted success rates, is not as accurate for poor candidates. This study examines the calculator's validity in an understudied multiracial cohort with a high proportion of poor candidates. Methods This retrospective study examined women with one or two prior cesarean deliveries who attempted VBAC at a single institution. Subjects were placed into quartiles based on MFMU-predicted success rates. For each quartile, actual and predicted success rates were compared. The calculated area under the receiver operating characteristic curve (AUC) was compared to the original AUC. Results The study included 1604 women. Actual and predicted VBAC rates were similar in the lowest and highest quartile groups, 18.2% vs. 21.2% (n = 11, P > 0.99) and 87.1% vs. 88.5% (n = 1090, P = 0.14), respectively. In the 51-75% predicted success rate group, the actual VBAC rate was higher than the predicted rate, 69.9% vs. 65.5% (n = 394) but not statistically significant (P = 0.07). In the 25-50% predicted success rate group, the actual VBAC rate was significantly higher than the predicted rate 55.1% vs. 39.6% (n = 109, P = 0.002). The actual AUC was lower than the MFMU model, 0.72 [95% confidence interval (CI) 0.69-0.75] vs. 0.77 (95% CI 0.76-0.78) (P < 0.001). Conclusion The MFMU VBAC calculator's predicted success rates were comparable to actual success rates for candidates with predicted success rates >75%. As predicted success rates declined, the calculator was increasingly inaccurate and underestimated the success rate. Caution should be taken when using the MFMU VBAC calculator for poor candidates.