Original Research

Predicting Failure of a Vaginal Birth Attempt

After Cesarean Delivery
Sindhu K. Srinivas, MD, David M. Stamilio, MD, MSCE, Erika J. Stevens, MA, Anthony O. Odibo, MD,
Jeffrey F. Peipert, MD, and George A. Macones, MD, MSCE
OBJECTIVE: To identify a group of clinical factors that
could be used to accurately predict failure in women
attempting vaginal birth after cesarean (VBAC).
METHODS: Weconducteda plannedsecondary analysis of a
retrospective cohort study of women whowere offeredVBAC
from 1996 to 2000 in 17 community and university hospitals.
We collected information about maternal history and out-
comes of the index pregnancy. We used univariable and
multivariable statistical methods to develop a multivariable
prediction model for the outcome of VBAC failure.
RESULTS: A total of 13,706 patients attempted VBAC, with a
failure rate of 24.5%. Six variables were significantly associated
with VBAC failure in our final logistic regression model: gesta-
tional age at delivery, maternal age, maternal race, labor type
(spontaneous, augmented, or induced), history of vaginal de-
livery, and cephalopelvic disproportion or failed induction
(combined variable) as prior cesarean indication. The area
under the receiver operating characteristics curve is 0.717. To
achieve a sensitivity of approximately 75%, a false-positive rate
of approximately 40% would result.
CONCLUSION: Our results indicate that significant clin-
ical variables (prelabor and labor) cannot reliably predict
VBAC failure.
(Obstet Gynecol 2007;109:8005)
LEVEL OF EVIDENCE: II
A
pproximately 26% of the 4 million births per year
in the United States are cesarean deliveries, and
this number will only continue to grow as the role of
elective primary cesarean delivery is further expand-
ed.
1
Over the past 2 decades, obstetricians have been
encouraged to offer women with a previous low
transverse cesarean an attempt at vaginal birth. Past
research has indicated that a strategy of vaginal birth
after cesarean delivery (VBAC) has a reasonable
success rateas high as 6080% in the overall VBAC
population.
2
However, these success rates differ based
on clinical factors. More recent research has focused
on maternal and neonatal risks associated with
VBAC, as well as predictors of VBAC success and
failure.
35
Complication rates appear to be greatest in
those who fail a trial of labor and subsequently
require a cesarean delivery.
6
Conversely, safety is
greatest in those who attempt a trial of labor after a
prior cesarean delivery and are successful.
Several studies have found that many factors are
associated with VBAC failure, such as preeclampsia,
macrosomia, maternal obesity, and labor induc-
tion.
3,4,7
Some have proposed different scoring mech-
anisms to incorporate these individual factors and
attempt to reliably predict VBAC success, failure, or
complications. Troyer and Parisi,
8
Vinueza et al,
9
and
Gonen and colleagues
10
have all attempted to develop
different scoring systems to predict VBAC success.
However, these systems were all developed using a
relatively small number of patients and do not incor-
porate several factors that have been associated with
VBAC success. In general, prior attempts to develop
a means of predicting a successful VBAC trial have
not been successful.
68
Our aim was to identify a group of clinical factors,
both prelabor and intrapartum, in a large multicenter
cohort that could be used to accurately predict failure
of achieving a vaginal delivery in women attempting
VBAC.
See related editorial on page 796.
From the Departments of Obstetrics and Gynecology, University of Pennsylvania
Health System, Philadelphia, Pennsylvania, and Washington University, St.
Louis, Missouri.
This study was supported by a grant from the National Institute of Child Health
and Human Development (NICHD) (RO1HD 35631 to G.A.M.). Dr. Macones
is a recipient of a Mid-Career Award in Patient-Oriented Research (K24 HD
4537328), which partially supported this work, and Dr. Peipert is a recipient
of a K24 grant from NICHD (K24 HD01298), which partially supports this
work.
Corresponding author: Sindhu K. Srinivas, MD, University of Pennsylvania
Medical Center, 2000 Courtyard Building, 3400 Spruce Street, Philadelphia,
PA 19104; e-mail: ssrinivas@obgyn.upenn.edu.
2007 by The American College of Obstetricians and Gynecologists. Published
by Lippincott Williams & Wilkins.
ISSN: 0029-7844/07
800 VOL. 109, NO. 4, APRIL 2007 OBSTETRICS & GYNECOLOGY
MATERIALS AND METHODS
We conducted a planned secondary analysis within a
retrospective cohort study of 25,005 women with
prior cesarean deliveries from 1996 to 2000. The
details of the study methods have been published
previously,
11,12
but a brief description follows. Of the
17 northeastern United States hospitals included in
the study, six were tertiary care centers with residency
programs, five were community-based hospitals with
residency programs, and six were nonteaching com-
munity hospitals. Individual hospital names are avail-
able upon request. All local institutional review
boards approved the conduct of the study. The cohort
of patients was identified via International Classifica-
tion of Diseases, 9th Revision (ICD-9) codes for prior
cesarean delivery, delivered. All charts for eligible
patients were sequentially studied. We excluded pa-
tients with unknown prior cesarean scar type or prior
classic uterine incision. There were 478 patients ex-
cluded for a prior vertical cesarean, and of these, 460
were termed classic and 18 just vertical in the medical
record. Trained research nurse abstractors collected
demographic information, medical and obstetric his-
tory, and data on complications, treatment, and out-
come of the index pregnancy using standardized,
closed-ended, data collection forms. To assure accu-
rate and valid data collection, training for abstractors
was repeated at various predetermined points during
the study and when quality assurance measures (re-
peat abstraction of 3%) indicated the need.
In the original retrospective cohort study, the
primary delivery-related adverse outcome was uterine
rupture. This outcome was identified by the data
abstractors using strictly defined diagnostic criteria
that have been previously described.
11,12
Secondary
adverse maternal outcomes included in the initial
cohort study were operative complications such as
bladder, ureter, or bowel injury and uterine artery
laceration.
For this planned secondary analysis, we identified
VBAC failure as the primary outcome of interest and
defined it as the inability to achieve a vaginal delivery
with a VBAC attempt. We used univariable and
multivariable (ie, logistic regression) statistical meth-
ods to develop a multivariable prediction model for
the outcome of VBAC failure. We selected potential
predictors for inclusion in the final regression model
based on the unadjusted analysis (P.2) and proven
or suspected biological importance. The final regres-
sion model was developed by sequentially removing
covariates and testing differences in hierarchical mod-
els with the likelihood ratio test; the final model
includes significant variables identified to be the most
predictive of VBAC failure. We also assessed hospital
as a confounder in this analysis and analyzed the
possibility of clustering effects by hospital in prior
analyses (Srinivas SK, Stamilio DM, Sammel MD,
Stevens EJ, Peipert JF Odibo AO, Macones GA.
Vaginal birth after cesarean delivery: does maternal
age affect safety and success? Paediatr Perinat Epide-
miol [in press]). The entire analysis was performed
with Stata 8 Special Edition (StataCorp, College Sta-
tion, TX).
RESULTS
The total number of patients included in the cohort
was 25,005. Of these, 13,706 (54.81%) attempted
vaginal birth after cesarean delivery, and 11,299
(45.19%) elected repeat cesarean. Of those who at-
tempted VBAC, there was a 24.5% failure rate.
The demographics of the entire cohort, as illus-
trated in Table 1, are representative of a general
obstetric population. The women who attempted a
vaginal birth were similar to those who elected a
cesarean delivery with respect to several factors, such
as age, gestational age at delivery, and birth weight
Table 1. Demographics of the Entire Cohort
Characteristic
VBAC
Attempt
(n13,706)
Cesarean
Delivery
(n11,299)
Race
White 7,553 (57) 7,329 (67)
Black 4,075 (31) 2,481 (23)
Other 1,659 (12) 1,120 (10)
Age (y)
1520 699 (5) 222 (2)
2134 9,801 (72) 7,573 (67)
35 or older 3,206 (23) 3,504 (31)
Insurance
Medicaid 3,221 (24) 1,852 (17)
HMO 4,742 (35) 4,072 (37)
Private 3,654 (27) 3,671 (33)
Other 1,779 (13) 1,384 (13)
University 6,234 (46) 3,980 (35)
Birth weight over 4,000 g 3,444 (25) 3,005 (27)
Gestational age (wk)
Less than 32 354 (3) 330 (3)
3237 850 (6) 961 (9)
3741 10,222 (75) 8,446 (75)
More than 41 2,280 (17) 1,562 (14)
Labor
Spontaneous 4,750 (35)
Augmented 4,696 (35)
Induced 3,988 (30)
VBAC, vaginal birth after cesarean; HMO, health maintenance
organization.
Data are expressed as n (%).
VOL. 109, NO. 4, APRIL 2007 Srinivas et al Predicting VBAC Failure 801
more than 4,000 g. In the VBAC and elective cesar-
ean groups, roughly two thirds of women were aged
2134 years. The rates of preterm birth in the VBAC
and elective cesarean groups were similar to each
other and to the national rate. Health care third party
payor types were evenly distributed and similar be-
tween the VBAC and elective cesarean groups. How-
ever, there were more African-American women and
women obtaining care at a university hospital in the
VBAC attempt group.
The unadjusted analysis is displayed in Table 2.
Several factors, including gestational age, the pres-
ence of diabetes or chronic hypertension, birth
weight, labor type, prior indication for cesarean de-
livery, and prior vaginal delivery, were all signifi-
cantly associated, either positively or negatively, with
failure of a VBAC attempt.
Table 3 exhibits the results of the multivariable
logistic regression analysis for the outcome, failed
VBAC. The final model consists of six variables,
including maternal race and age, gestational age at
delivery, spontaneous (versus induced) labor, prior
cesarean indication of cephalopelvic disproportion or
failed induction, and history of vaginal delivery. The
variables diabetes, chronic hypertension, and birth
weight more than 4,000 g were not included in the
final predictive model because, although they were
associated with VBAC failure, they did not improve
predictive accuracy and were felt to be biologically
less important for prediction of VBAC failure because
of modest effect size or prevalence or both. Having a
prior vaginal delivery was the most protective against
a failed VBAC attempt (odds ratio 0.21, 95% confi-
dence interval 0.190.24), and labor induction had
the largest magnitude of association with VBAC
failure. Figure 1 illustrates the receiver operating
characteristics curve for our prediction model. The
area under the receiver operating characteristics
curve is 0.7170. This model for VBAC failure predic-
tion has a sensitivity of 11.08%, a specificity of
96.64%, a positive predictive value of 52.17%, and a
negative predictive value of 76.66%. The receiver
operating characteristics curve reveals an unfavorable
trade-off between sensitivity and false-positive rate for
this prediction model. For example, selecting an
optimal cutoff point on the curve to achieve 75%
Table 2. Failure Compared With Success of Vaginal Birth After Cesarean Delivery (Bivariate Analysis)
Characteristic
VBAC Failure
(n3,366)
VBAC Success
(n10,340)
OR
(95% CI) P
Race
White 1,819 (54) 5,734 (56) Reference Reference
Black 1,077 (32) 2,998 (30) 1.13 (1.041.24) .005
Hispanic 169 (5) 556 (1) 0.96 (0.801.15) .642
Asian 79 (3) 225 (0.2) 1.11 (0.851.44) .447
Other 149 (5) 481 (0.5) 0.98 (0.811.18) .807
Age (y)
1520 148 (4) 551 (5) 0.82 (0.680.99) .038
2134 2,417 (72) 7,384 (71) Reference Reference
35 or more 801 (24) 2,405 (23) 1.02 (0.931.12) .712
Birth weight more than 4,000 g 937 (28) 2,507 (24) 1.15 (1.081.22) .001
Gestational age (wk)
Less than 32 44 (1) 310 (3) 0.40 (0.290.54) .001
3237 159 (5) 691 (7) 0.64 (0.540.77) .001
3741 2,697 (80) 7,525 (73) Reference Reference
More than 41 466 (14) 1,814 (18) 0.72 (0.640.80) .001
Gestational diabetes 197 (6) 414 (4) 1.46 (1.241.72) .001
Chronic hypertension 124 (4) 266 (3) 1.43 (1.161.77) .001
Labor type
Spontaneous 761 (23) 3,989 (39) Reference Reference
Augmented 1,209 (37) 3,487 (34) 2.68 (2.422.97) .001
Induced 1,349 (41) 2,639 (26) 1.81 (1.642.01) .001
Number of prior cesarean deliveries more than 1 293 (9) 878 (9) 1.03 (0.901.16) .71
Prior cesarean indication
Any CPD (failed IOL, CPD) 924 (27) 1,863 (18) 1.72 (1.571.89) .001
Prior vaginal delivery 507 (15) 4,534 (44) 0.34 (0.320.37) .001
VBAC, vaginal birth after cesarean delivery; OR, odds ratio; CI, confidence interval; CPD, cephalopelvic disproportion; IOL, induction
of labor.
Data are expressed as n (%).
802 Srinivas et al Predicting VBAC Failure OBSTETRICS & GYNECOLOGY
sensitivity for VBAC failure results in a false-positive
rate of about 40%.
When evaluating subgroups of women by using
the two strongest clinical factors, namely, labor type
and prior vaginal delivery history, the unadjusted rate
of VBAC success was highest for women with a prior
vaginal delivery and spontaneous labor (Table 4).
Conversely, women without a prior vaginal delivery
who underwent labor induction had the lowest
VBAC success rate. Table 4 summarizes the adjusted
odds of VBAC failure and observed rates of VBAC
success for particular subsets of women using the two
strongest clinical risk factors.
DISCUSSION
In our large retrospective cohort, VBAC failure can-
not be predicted efficiently by using a combination of
the strongest clinical risk factors. However, there
appear to be strong factors that are highly associated
with VBAC failure (ie, labor induction) and VBAC
success (ie, prior vaginal delivery). The presence or
absence of these factors may help in counseling
patients about attempting a trial of labor.
With the rise in cesarean delivery rate, there has
been a growing interest in identifying long-term or
down-stream reproductive health consequences of a
continually increasing number of cesarean deliveries.
Concerns include increased maternal operative com-
plications and adverse neonatal outcomes, particu-
larly in patients with multiple prior cesareans and in
patients who require cesarean delivery after a trial of
labor. In an effort to better target therapy and mini-
mize further inflation of the rates of cesarean delivery
and adverse perinatal and maternal outcome, clinical
researchers have sought reliable predictors of VBAC
success.
Maternal obesity, pregestational diabetes, in-
creasing number of prior cesarean deliveries, fetal
Table 3. Final Logistic Regression Model for Prediction of Failure of Vaginal Birth After Cesarean
Delivery
Adjusted OR 95% CI P
Race
White Reference Reference Reference
Black 1.72 1.561.91 .001
Hispanic 1.45 1.191.77 .001
Asian 1.25 0.951.65 .108
Other 1.16 0.941.42 .161
Age (y)
1520 0.56 0.450.68 .001
2134 Reference Reference Reference
35 or more 1.22 1.101.35 .001
Gestational age (wk)
Less than 32 0.43 0.300.60 .001
3237 0.73 0.600.88 .001
3741 Reference Reference Reference
More than 41 0.80 0.710.91 .001
Labor type
Spontaneous Reference Reference Reference
Augmented 1.53 1.371.70 .001
Induced 2.39 2.152.67 .001
Prior cesarean indication
Any CPD (failed IOL, CPD) 1.44 1.301.58 .001
Prior vaginal delivery 0.21 0.190.24 .001
OR, odds ratio; CI, confidence interval; CPD, cephalopelvic disproportion; IOL, induction of labor.
Fig. 1. Prediction of failure of vaginal birth after cesarean
delivery. Multivariable model (area under the receiver
operating characteristics curve0.7170).
Srinivas. Predicting VBAC Failure. Obstet Gynecol 2007.
VOL. 109, NO. 4, APRIL 2007 Srinivas et al Predicting VBAC Failure 803
macrosomia, and having had a prior cesarean for
cephalopelvic disproportion decrease the chance of
VBAC success, whereas nonrecurring or nonpersis-
tent indications for cesarean (ie, breech presentation
and nonreassuring fetal heart tracing) are associated
with an increased likelihood of success.
3,5,6,13
Having
had a prior vaginal delivery is strongly associated with
increased VBAC success.
14
Our results corroborate
this finding.
The study design used for this investigation, a
multicenter retrospective cohort, has several impor-
tant advantages. First, the data set is robust, with
information on clinically relevant exposure, outcome,
and confounding variables that were strictly defined
with a minimal amount of missing data. Second, the
results are more generalizable to many obstetric pop-
ulations because subjects were selected from various
types of hospitals. Third, the rates of uterine rupture
and other adverse VBAC-related outcomes in our
cohort corroborate those estimated in other studies,
imparting validity to our results.
2,1518
Although this study has many attributes, there are
some limitations. First, we were restricted to the
variables contained in the data set. Although the data
set is fairly comprehensive, some covariates are not
available. For example, we had no data on maternal
weight, body mass index, weight gain during preg-
nancy, or detailed characteristics of labor induction
techniques. These variables may be biologically im-
portant predictors of VBAC failure, especially mater-
nal obesity. However, given the previously observed
modest magnitude of effect of obesity on vaginal
delivery rate, it is unlikely that this one factor would
appreciably improve our prediction model. Addition-
ally, we had no information on neonatal outcomes
because this study was primarily established to assess
maternal outcomes of VBAC. Second, if the clinicians
had preconceptions regarding VBAC and success that
systematically affected their counseling techniques,
there is the potential for confounding by indication.
This study specifically evaluates the effect of
various clinical factors (prelabor and labor) on VBAC
failure and attempts to develop a prediction rule for
failed VBAC. These data validate prior studies that
identified several risk factors associated with VBAC
failure and protective factors associated with success.
Despite our large multicenter cohort and rigorous
analysis, we were unable to develop a single reliable
clinical prediction rule for VBAC failure that can be
applied to all women. However, we were able to
identify clinically helpful information for VBAC
counseling using the two strongest clinical risk factors,
prior vaginal delivery and labor type. Obstetricians
can modify their counseling about trial of labor using
the knowledge about known factors associated with
VBAC failure and success found in our analysis.
REFERENCES
1. Martin JA, Hamilton BE, Sutton PD, Ventura SJ, Menacker F,
Munson ML. Births: final data for 2002. Natl Vital Stat Rep
2003;52:1113.
2. Rosen MG, Dickinson JC, Westhoff CL. Vaginal birth after
cesarean: a meta-analysis of morbidity and mortality. Obstet
Gynecol 1991;77:46570.
3. Landon MB, Leindecker S, Spong CY, Hauth JC, Bloom S,
Varner MW, et al. The MFMU Cesarean Registry: factors
affecting the success of trial of labor after previous cesarean
delivery. Am J Obstet Gynecol 2005;193:101623.
4. Landon MB, Hauth JC, Leveno KJ, Spong CY, Leindecker S,
Varner MW et al. Maternal and Perinatal outcomes associated
with a trial of labor after prior cesarean delivery. N Engl J Med
2004;351:25819.
5. Brill Y, Windrim R. Vaginal birth after Caesarean section:
review of antenatal predictors of success. J Obstet Gynaecol
Can 2003;25:27586.
Table 4. Odds of VBAC Failure and Observed VBAC Success Rates for Subgroups of Patients Identified
by Risk Factor Combinations of Labor Type*
Adjusted OR (95% CI)

P VBAC Success Rate (%)

No prior vaginal delivery and induced labor Reference Reference 57

Prior vaginal delivery
Spontaneous labor 0.14 (0.120.16) .001 94
Augmented labor 0.34 (0.290.39) .001 89
Induced labor 0.21 (0.190.24) .001 84
No prior vaginal delivery
Spontaneous labor 0.65 (0.590.73) .001 75
Augmented labor 1.56 (1.421.73) .83 67
OR, odds ratio; CI, confidence interval; VBAC, vaginal birth after cesarean delivery.
* Spontaneous, augmented, or induced and prior vaginal delivery history (yes/no).

Multifactor odds ratios were calculated as linear combinations using the predictive model.

An observed unadjusted rate.

804 Srinivas et al Predicting VBAC Failure OBSTETRICS & GYNECOLOGY
6. McMahon M, Luther ER, Bowes WA, Olshan AF. Compari-
son of a trial of labor with an elective second cesarean section.
N Engl J Med 1996;335:68995.
7. Carroll C, Magann EF, Chauhan SP, Klauser CK, Morrison
JC. Vaginal birth after cesarean section versus elective repeat
cesarean delivery: Weight-based outcomes. Am J Obstet
Gynecol 2003;188:151622.
8. Troyer L, Parisi V. Obstetric parameters affecting success in a
trial of labor: designation of a scoring system. Am J Obstet
Gynecol 1992;167:1099104.
9. Vinueza CA, Chauhan SP, Barker L, Hendrix NW, Scardo JA.
Predicting the success of a trial of labor with a simple scoring
system. J Reprod Med 2000;45:3326.
10. Gonen R, Tamir A, Degani S, Ohel G. Variables associated
with successful vaginal birth after one cesarean section: a
proposed vaginal birth after cesarean section score. Am J
Perinatol 2004;21:44753.
11. Macones GA, Cahill A, Pare E, Stamilio DM, Ratcliffe S,
Stevens E, et al. Obstetric outcomes in women with two prior
cesarean deliveries: is vaginal birth after cesarean delivery a
viable option? Am J Obstet Gynecol 2005;192:12239.
12. Macones GA, Peipert J, Nelson DB, Odibo A, Stevens EJ,
Stamilio D, et al. Maternal complications with vaginal birth
after cesarean delivery: a multicenter study. Am J Obstet
Gynecol 2005;193:165662.
13. Elkousy MA, Sammel M, Stevens E, Peipert JF, Macones G.
The effect of birth weight on vaginal birth after cesarean
delivery success rates. Am J Obstet Gynecol 2003;188:
82430.
14. Bedoya C, Bartha JL, Rodriguez I, Fontan I, Bedoya JM,
Sanchez-Ramos J. A trial of labor after cesarean section in
patients with or without a prior vaginal delivery. Int J Gynaecol
Obstet 1992;39:2859.
15. Pickhardt MG, Martin JN, Meydrech EF, Blake PG, Martin
RW, Perry KG, et al. Vaginal birth after cesarean delivery: are
there useful and valid predictors of success or failure? Am J
Obstet Gynecol 1992;166:18119.
16. Farmer R, Kirschbaum T, Potter D, Strong TH, Medearis AL.
Uterine rupture during trial of labor after previous cesarean
section. Am J Obstet Gynecol 1991;165:9961001.
17. Jones RO, Nagashima AW, Hartnett-Goodman MM, Goodlin
RC. Rupture of low transverse cesarean scars during trial of
labor. Obstet Gynecol 1991;77:8157.
18. Scott JR. Avoiding labor problems during vaginal birth after
cesarean delivery. Clin Obstet Gynecol 1997;40:53341.
HighWire Press
Streamline Your Search
Searching content in the leading scientific journals is now easier with the HighWire Portal.
With over 1 million free full-text articles, HighWire is host to the largest archive of free bio-
medical research in the world. Free access is also provided to the full text of cited references
in all HighWire-hosted journals. By registering with the portal, users can quicklyviewwhich
articles in their search results (from over 130 scholarly publishers) are available for free, by
current subscription, or through pay-per-view.
More Content: Includes over 15 million articles in more than 4,500 MEDLINE journals.
Better Searching: Search across the full-text of all 1,011 HighWire-hosted online journals,
plus the entire MEDLINE database, by author, keyword, or citation. Tools for discovery
include: Concept/topic browsing using HighWire's emerging taxonomy, keyword in context
display, "instant index" of clustered search results, citation mapping (showing the most
highly-cited articles directly related to an article), and a weighted topic-matching tool.
More Alerting: Sign up to receive Tables of Contents and new content alerts matching key-
words, authors, citations, and topics in any of the HighWire-hosted titles plus all of
MEDLINE.
Find just what you need at www.highwire.org
VOL. 109, NO. 4, APRIL 2007 Srinivas et al Predicting VBAC Failure 805