A prediction model of vaginal birth after cesarean in the preterm period




Materials and Methods


We performed a secondary analysis of a prospective cohort study designed to evaluate perinatal outcomes in women with a prior uterine scar, using information obtained from 19 academic medical centers belonging to the Eunice Kennedy Shriver National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network between 1999 and 2002.


Women with a prior cesarean delivery and a current singleton pregnancy at 20 weeks or more or whose infant had a birthweight of at least 500 g were included in the original cohort study. Women were classified as not undergoing a trial of labor if they underwent an elective repeat cesarean delivery without labor or there were other indications for cesarean delivery, including a prior classical or T incision, malpresentation, placenta previa, prior myomectomy, nonreassuring fetal heart tracing, genital herpes, or a medical condition precluding a trial of labor.


In this study, we included women with 1 prior low transverse cesarean delivery undergoing a trial of labor after cesarean delivery with a vertex singleton pregnancy in the preterm period (26 0/7 weeks to 36 6/7 weeks). Women were excluded if they had an in utero fetal demise, more than 1 prior cesarean delivery, a prior cesarean delivery with a non–low transverse or an unknown scar, were missing gestational age information, or did not attempt a trial of labor after cesarean delivery.


We collected variables thought to modify the likelihood of vaginal birth after cesarean delivery success, including maternal age, race/ethnicity, prior vaginal deliveries, pregestational diabetes, chronic hypertension, prior vaginal birth after cesarean delivery, recurring indication for prior cesarean delivery (defined as cephalopelvic disproportion, failure to progress, or failed induction), and factors known on admission for delivery: gestational age, cervical examination, body mass index (kilograms per square meter), gestational diabetes, gestational hypertension or preeclampsia, and induction of labor. Estimated fetal weight was not included because the information was not available; regardless, ultrasound estimation of fetal weight is known to be imprecise, especially in obese patients.


Statistical analysis


We first examined the associations between maternal sociodemographic, obstetrical complications, and characteristics on admission to labor and delivery in relation to the success of vaginal birth after cesarean delivery. Categorical variables were analyzed using either the χ 2 or Fisher exact probability tests; continuous variables were analyzed based on the Student t test. For continuous variables that were not normally distributed, we applied the 2-sample median test for comparisons. All risk factors that reached significance at P < .1 were considered as potential candidates for inclusion in the multivariable prediction models.


We developed a prediction model for vaginal birth after cesarean delivery success based on a logistic regression. Prior to developing this prediction model, we created a random 70% to 30% split of the original data. All prediction models were first developed in the 70% data (training sample), and the final model was subsequently validated in the 30% data (validation sample).


In the training data set, we developed the prediction model by including all risk factors from the univariable analysis (that met the criteria of P < .1); these included maternal age, parity, body mass index at delivery, prior vaginal delivery, chronic hypertension, hypertensive disorder of pregnancy (including gestational hypertension or preeclampsia), diabetes mellitus (pregestational or gestational), admission cervical examination, and labor induction. We also considered potential 2-way interactions among risk factors and retained interaction terms with P < .15 based on the likelihood ratio test of nested models with and without the risk factor in question. The fit of the logistic regression model was assessed based on the Hosmer-Lemeshow goodness-of-fit test.


Once the final model was derived, we validated the model in the validation data set. From this validated model, we calculated the area under the curve with a 95% confidence interval. Given the small size of the validation cohort, we estimated the variance (and hence the 95% confidence interval) of the area under the curve through 1000 bootstrap resamples with replacement. Finally, to allow for overoptimistic predictions on the area under the curve, we recalculated the area under the curve after applying a 10-fold jackknife cross-validation to the sample. All statistical analyses were carried out using SAS version 9.4 (SAS Institute, Cary, NC).




Results


Among the data set, 1295 women met our criteria for analysis. Of these women, 76.6% had a successful vaginal birth after cesarean delivery ( Figure 1 ). Descriptive characteristics of the patients used in our model are presented in Table 1 . In the Supplemental Table , we provide descriptive characteristics of patients in the training dataset compared to the validation dataset. The results of the logistic regression model are shown in Table 2 . The factors that significantly decreased the rate of vaginal birth after cesarean delivery success were chronic hypertension (odds ratio, 0.56, 95% confidence interval, 0.26–1.20) and recurring indication in prior cesarean (odds ratio, 0.63, 95% confidence interval, 0.41–0.97). The factors that significantly increased the rate of vaginal birth after cesarean delivery success were a prior vaginal delivery (odds ratio, 1.75, 95% confidence interval, 1.02–3.01) and a prior vaginal birth after cesarean delivery (odds ratio, 1.78, 95% confidence interval, 0.94–3.37).




Figure 1


Flow chart of patients undergoing a TOLAC in the preterm period after applying exclusion criteria

Flow chart of the study population and outcome of trial of labor after cesarean delivery is shown.

IUFD , intrauterine fetal demise; TOLAC , trial of labor after cesarean delivery.

Mardy et al. Preterm VBAC calculator. Am J Obstet Gynecol 2016 .


Table 1

Descriptive characteristics of preterm patients undergoing a TOLAC


















































































































Admission variable Failed TOLAC (n = 303) Successful TOLAC (n = 992) P value
Maternal age, mean y (SD) 28.5 (5.9) 28.3 (6.1) .712
BMI, mean (SD) 32.7 (8.2) 30.4 (6.7) < .001
Parity, median (range) 1 (1-8) 2 (1–12) < .001
Cervical dilation, median, cm (range) 1 (0–9) 3 (0–10) < .001
Cervical effacement, median, % (range) 50 (0, 100) 75 (0–100) < .001
Cervical station, median (range) –3 (floating/ballotable, 0) –2 (floating/ballotable, +3) < .001
Gestational age, mean (SD) 33 wks 4 d (2 wks 4 d) 33 wks 6 d (2 wks 3 d) .116
Mean birthweight, g (SD) 2162 (781.5) 2308 (646) .003
Ethnicity .650
White, n, % 125 (41.3) 371 (37.4)
Black, n, % 128 (42.2) 456 (46.0)
Hispanic, n, % 39 (12.9) 127 (12.8)
Other, n, % 11 (3.6) 38 (3.8)
Smoking 62 (20.5) 264 (26.7) .030
Diabetes, n, % a 45 (14.9) 86 (8.7) .002
cHTN, n, % 37 (12.2) 35 (3.5) < .001
Hypertensive disease of pregnancy, n, % b 111 (36.6) 99 (10.0) < .001
Induction of labor, n, % 134 (44.2) 225 (22.7) < .001
Recurring indication in prior cesarean, n, % c 85 (28.1) 239 (24) .164
Any prior vaginal delivery 115 (38.6) 581 (59.2) < .001
Prior VBAC 53 (18.7) 357 (38.4) < .001

BMI , body mass index; cHTN , chronic hypertension; TOLAC , trial of labor after cesarean delivery; VBAC , vaginal birth after cesarean.

Mardy et al. Preterm VBAC calculator. Am J Obstet Gynecol 2016 .

a Pregestational or gestational diabetes


b Gestational hypertension or preeclampsia


c Indication for previous cesarean delivery is listed as cephalopelvic disproportion, failure to progress, or failed induction.



Table 2

Logistic regression model for predicting the success of vaginal birth after a cesarean delivery in preterm gestations































































































Parameter Coefficient (β) SE (β) P -value Odds ratio 95% Confidence interval
Intercept 0.934 0.232 <0.001
Chronic hypertension –0.601 0.391 0.124 0.55 0.26–1.18
Any prior vaginal delivery 0.420 0.245 0.087 1.52 0.94–2.46
Prior VBAC 0.715 0.302 0.018 2.04 1.13–3.69
Recurring indication in prior cesarean delivery –0.466 0.219 0.034 1.52 0.41–0.96
Cervical dilation on admission 0.340 0.071 <0.001
HDP –2.325 0.595 <0.001
IOL –0.729 0.260 0.005
Diabetes a 0.138 0.394 0.727
HDP × IOL 1.583 0.605 0.009
Cervical dilation, cm, × HDP 0.481 0.204 0.018
Cervical dilation, cm, × diabetes a –0.420 0.150 0.005

HDP , hypertensive disease of pregnancy (gestational hypertension or preeclampsia); IOL , induction of labor; VBAC , vaginal birth after cesarean.

Mardy et al. Preterm VBAC calculator. Am J Obstet Gynecol 2016 .

a Pregestational or gestational.



Other variables were included in the model in 2-way interactions. These variables included cervical dilation on admission, hypertensive disease of pregnancy (preeclampsia and gestational hypertension), induction of labor, and diabetes (pregestational and gestational). Of note, because these variables were included in the 2-way interactions, odds ratios are not provided.


Cervical dilation is associated with increased vaginal birth after cesarean delivery success alone and when associated with hypertensive disease of pregnancy; the rate decreases if diabetes is present. In a patient with hypertensive disease of pregnancy, the variable alone is strongly associated with a decreased rate of vaginal birth after cesarean delivery success, but it increases if a patient is induced or cervical dilation is present. For women who require an induction of labor, this is negatively correlated with vaginal birth after cesarean delivery success, but it improves if hypertensive disease of pregnancy is present. Finally, for women with diabetes, there is a decreased chance of success the more dilated a patient is.


The area under the curve in the validation model is shown in Figure 2 . The overall area under the curve of the model was 0.80 (95% confidence interval, 0.77–0.83). The area under the curve from the model based on the bootstrap resampling was 0.78 (95% confidence interval, 0.74–0.82), and the model fit the data well (Hosmer-Lemeshow, P = .367). Following application of the 10-fold cross-validation jackknife methodology, the corrected area under the curve was similar to the bootstrap estimate (0.77, 95% confidence interval, 0.73–0.82). The logistic regression coefficients for each variable in the model are shown in Table 2 .




Figure 2


ROC for prediction of VBAC in preterm after trial of labor

The ROC curve for the multivariable logistic regression model for the prediction of VBAC in the preterm period after a trial of labor is shown. The area under the curve was 0.80 (95% confidence interval, 77–83).

ROC , receiver-operating characteristic; VBAC , vaginal birth after cesarean.

Mardy et al. Preterm VBAC calculator. Am J Obstet Gynecol 2016 .


The prediction probabilities for 4 sample cases are shown in Table 3 . These represent the results of the logistic regression equation (see the logistic regression equation in the following text) based on hypothetical patient variables and the predicted chance of vaginal birth after cesarean delivery with corresponding 95% confidence intervals.



Table 3

Prediction of VBAC success for 4 hypothetical patients
































































Variable Case 1 Case 2 Case 3 Case 4
Chronic hypertension Yes Yes No No
Any prior vaginal delivery No Yes Yes Yes
Prior VBAC No No Yes Yes
Recurring indication for prior cesarean Yes Yes No No
Dilation on admission, cm 0 1 2 4
Hypertensive disease of pregnancy Yes No Yes No
Induction of labor Yes Yes No No
Pregestational or gestational diabetes Yes Yes Yes No
Predicted VBAC success (95% confidence interval) 19% (7.5–39.6%) 41% (19.9–65.2%) 67% (41.4–84.8%) 97% (94.6–98.2%)

Only gold members can continue reading. Log In or Register to continue

Stay updated, free articles. Join our Telegram channel

May 2, 2017 | Posted by in GYNECOLOGY | Comments Off on A prediction model of vaginal birth after cesarean in the preterm period

Full access? Get Clinical Tree

Get Clinical Tree app for offline access