Development and validation of an improved prediction model for vaginal birth after previous cesarean section: a retrospective cohort study of a Chinese population

Abstract

Objective: The transition from one-child to two-child and three-child policy in China has increasingly led to a rise in the number of women who choose trial of labor after cesarean section (TOLAC). Achieving vaginal birth after cesarean section (VBAC) is, however, not always guaranteed, and a failed TOLAC is associated with a high risk of maternal and neonatal complications. Although Grobman's model may help predict VBAC, variations in population characteristics and healthcare settings can limit its generalizability and validity on a global scale. This study, therefore, seeks to develop and validate an improved prediction model for VBAC at the onset of labor among the Chinese population.

Methods: Seven hundred and twenty women who attempted a TOLAC were enrolled. The development dataset comprised 481 women, while the other 239 women constituted the temporal validation dataset. Variable selection was executed using the least absolute shrinkage and selection operator method. Model development was performed using logistic regression techniques and was presented as a nomogram.

Results: Of the participants, 81.4% achieved VBAC. The model included maternal age, maternal height, ratio of weight gain to pre-pregnancy weight, interval time of pregnancies, previous vaginal delivery, premature rupture of membranes, oxytocin administration, spontaneous labor onset, labor analgesia, and newborn weight. The development and temporally validated areas under the curve were 0.780 (95% confidence interval 0.726-0.834) and 0.774 (95% confidence interval 0.694-0.854), respectively. Internal validation performed by bootstrap resampling, calibration curves, and Hosmer-Lemeshow test confirmed the model's robust performance. An optimal predicted probability cut-off of 0.7 was identified by decision curve analysis and clinical considerations.

Conclusions: The improved predictive VBAC model exhibited adequate performance such that women with a prior low transverse cesarean delivery who scored 0.7 or higher (in the model-derived probability score) would consider TOLAC, potentially leading to a reduction in maternal-neonatal morbidity.

Registration: The study was approved by the Ethical Committee of Obstetrics and Gynecology Hospital, Fudan University (2018-43) and was registered in the Chinese Clinical Trial Registry (ChiCTR1900022484), https://www.chictr.org.cn/showproj.html?proj=37898. The study adhered to the Declaration of Helsinki. The first participant was enrolled on January 1, 2016. The requirement for informed consent was waived because the data were anonymized.

Keywords: Chinese population; Vaginal birth after cesarean; interval time of pregnancies; prediction model; ratio of weight gain to pre-pregnancy weight; temporal validation.

Figure 1.

Patients flow chart.

Figure 2.

Variables selection by least absolute shrinkage and selection operator (LASSO). (a) variables selection by LASSO regression model with the coefficient profiles of the 12 variables. (b) Variables selection by LASSO regression used 10-fold cross-validation while the tuning parameter λ was selected in the minimal average error.

Figure 3.

Nomogram, ROC plot, calibration plot and DCA plot of the model in development dataset. (a) a nomogram for the prediction of the likelihood of achieving vaginal birth after cesarean (VBAC). To use the nomogram, identify the position of every variable on its corresponding axis, draw a vertical line to the points axis to get the corresponding points of the variable, add all the points from every variable to the total points, and draw a vertical line from the total points axis to the axis representing ‘the success of TOLAC’ for the probability of achieving VBAC. (b) The area under the receiver-operator characteristic curve (ROC) of predicting successful VBAC among women with the trial of labor after cesarean section (TOLAC) in the development dataset. (c) Calibration plot with 500 bootstrap resamples for the internal validation to predict successful VBAC among women with TOLAC in the development dataset. The diagonal interrupted line is the line of ideal agreement. The red and the blue solid curves indicate a calibration curve using the loess nonparametric smoother and 500 bootstrap resamples, respectively. (d) Decision curve analysis (DCA) for the prediction model by 10-fold cross-validation in different thresholds in the development dataset. The red cross-validated curve indicated the average results was obtain from the 10-fold cross-validation and showed the robust estimates of the net benefit across all thresholds.

Figure 4.

ROC Plot and calibration plot of prediction model for achieving VBAC in temporal validation dataset. (a) The area under the receiver-operator characteristic curve(ROC)of predicting successful vaginal birth after cesarean (VBAC) among women with the trial of labor after cesarean section (TOLAC) in the temporal validation dataset. (b) Calibration plot with 500 bootstrap resamples for the temporal validation to predict successful VBAC among women with TOLAC in the temporal validation dataset. The diagonal interrupted line is the line of ideal agreement. The red and the blue solid curves indicate a calibration curve using the loess nonparametric smoother and 500 bootstrap resamples, respectively.