Chapter  71:  Vaginal Birth After Cesarean (VBAC) Volume 1 and Volume 2

General

• In the absence of RCTs of TOL versus repeat cesarean, evidence that is most generalizable comes from large country, State, or regional population-based studies (referred to as population-based studies) followed by large multicenter cohort studies, large single-institution or single-practice cohort studies, then smaller cohort studies, respectively.

• There is no direct evidence regarding the benefits and harms of TOL relative to ERCD in women who are similar in every respect except choice of delivery route.

• Several fair and good quality studies provide indirect evidence about relative benefits and harms of each route.

Maternal

• Maternal death rates did not differ between TOL and ERCD.

• The best evidence suggests that hysterectomy rates do not differ between TOL and ERCD.

• No studies examined specifically the risks of incontinence or pelvic support disorders in women with prior cesarean.

• Rates of infection were increased in ERCD versus TOL overall. Studies that performed subgroup analyses for TOL with and without vaginal delivery consistently found increased rates of infection for women who attempted TOL but ultimately had a cesarean delivery.

• There is conflicting evidence regarding whether induction of labor affects infection rates.

Infant

• There is insufficient evidence regarding the effect of selected route of delivery and Apgar score or respiratory morbidity.

• No study measured infant death directly attributable to a mother's choice of TOL or repeat CD.

• There is uncertainty about the magnitude of risk of perinatal death due to TOL. Results from two large studies differ in the magnitude of increased risk from TOL versus ERCD (90/1,000 TOL versus 50/1,000 ERCD compared with 12.9/1,000 TOL versus 1.1/1,000 ERCD). Neither study provides direct evidence of risk.

General

• Studies of legislation, policy, guidelines, hospital characteristics, provider characteristics, insurance type, or access to care focus exclusively on VBAC rates rather than safety.

Legal

• No study provides direct evidence for the impact of malpractice issues on VBAC or ERCD.

• One study reported that VBAC rates increased when legislation was enacted that standardized VBAC guidelines had to be provided to obstetric providers.

• The best evidence suggests that use of opinion leaders provides a greater likelihood of changing practice compared with audit and feedback.

Provider

• Studies of provider characteristics failed to control for important variables such as patient selection bias.

Hospital

• VBAC rates were higher in teaching hospitals compared to private, community, regional, or non-teaching hospitals.

• Three studies conflicted over the effect of hospitals containing a neonatal intensive care unit (NICU).

Insurance

• There was conflicting evidence regarding whether insurance status predicts VBAC.

Individual Factors

Data extraction and data entry were performed using Microsoft Excel 2000®. Because of the nature of this topic and the need for confounding consideration, further analysis involving the calculation of summary estimates using random effects modeling was not considered. Adjusted odds ratios (ORs) for the likelihood of VBAC from each study formed the basis for evaluation. In the situation where the study provided adjusted OR for the likelihood of a failed TOL, the inverse ratio was taken, to approximate the OR for the likelihood of VBAC.

Summary

• Rates of vaginal delivery when attempting TOL ranged from 60–82 percent. The largest population-based study reported a rate of 60.4 percent. The combined vaginal delivery rate for all prospective cohort studies, largely conducted in university or tertiary care settings, was 75.9 percent

• There was a 10 percent reduction in the likelihood of vaginal delivery when oxytocin was used for ether induction or augmentation. There was a similar trend for prostaglandins.

Scoring Systems

Table 2. Flamm Scoring System Tool: Included variables (more...)

Table 2. Flamm Scoring System Tool: Included variables and point values

Variable	Beta Coefficient	Point Value
Age under 40 years	0.95	2
Vaginal birth history
Before and after 1st cesarean	2.21	4
After 1st cesarean	1.22	2
Before 1st cesarean	0.43	1
None	Referent	0
Reason other than FTP for 1st cesarean	0.66	1
Cervical effacement at admission
> 75%	1.00	2
25%–75%	0.58	1
<25%	Referent
Cervical dilation 4cm or more at admission	0.77	1

Taken from Flamm, 1997³⁶

Table 3. Flamm Scoring System Tool: Performance of (more...)

Table 3. Flamm Scoring System Tool: Performance of Admission Score in the score testing group

Score	# of subjects with score	% of subjects with VBAC
0 to 2	114	49.1
3	329	59.9
4	595	66.7
5	660	77.0
6	360	88.6
7	189	92.6
8 to 10	158	94.9
Total	2405	74.9

Taken from Flamm, 1997³⁶

Table 4. Scoring System Tools: Relationship of risk (more...)

Table 4. Scoring System Tools: Relationship of risk score to successful VBAC

	Troyer, 199240			Vinueza, 200041
Score	Total # of subjects	% of subjects with VBAC	% False Positive/ Negative	Total # of subjects	% of subjects with VBAC	% False Positive/ Negative
0	59	91.5	2	56	98	0.4
1	92	73.9		106	69
2	87	66.7		74	40
3 to 4	26	46.1	5	27	33	3
Overall	264	74.9		263	63

Other scoring systems were developed retrospectively and have not been validated in a second sample.

Would these prediction tools be useful in practice? The probability that a woman would have a vaginal delivery is likely to influence her enthusiasm about trial of labor. Additionally, women who have a cesarean after a lengthy trial of labor are more likely to sustain adverse events such as uterine rupture or infection. Therefore, a tool that could accurately predict a woman's likelihood of achieving vaginal delivery with minimal adverse sequelae would be of interest to clinicians and patients. The value of a prediction tool depends on how it affects decisions about the likelihood of false positive and false negative tests (e.g., its accuracy), and the relative costs (harms) of false positive and/or negative results. The vaginal delivery rate in Flamm's population (e.g.. the overall rate of vaginal delivery), was 74.9 percent. Thirty percent of his population would be predicted to have a high probability of vaginal delivery (e.g., score or 6–10), and 18 percent were predicted to have a low likelihood of vaginal delivery (e.g., scores of 0–3). Slightly over half of the population would gain no additional information from using the predictive tool. Ten percent of the population or 253/2,405 may have been advised to have a cesarean, due to tool's prediction of low likelihood of vaginal delivery, when they would have been able to have a vaginal delivery. This may be acceptable as the harms of having a repeat cesarean may be low. What may be of higher concern is the false positive rate, or the chance that the tool would have encouraged TOL but the patient ended up with a cesarean. This is of higher concern because this group is of higher likelihood of sustaining complications from TOL such as infection and uterine rupture. This tool has a relatively low false positive rate of 2.6 percent (63/2405). Troyer's population had a similar vaginal delivery rate of 73 percent. The tool only provided additional information, to 32 percent of the population, with 22 percent predicted to have a high chance of vaginal delivery (e.g. score of 0), and 10 percent predicted to have a low chance (e.g. score of 3 or 4). This tool had a similar false positive rate of 2 percent (5/264), and slightly improved false negative rate at 4.5 percent (9/264). When this tool was used in a population with a lower pretest probability for vaginal delivery, both the false positive rate and false negative rate improved. Vinueza's population had a 63 percent vaginal delivery rate, 21 percent were predicted to have a high chance of vaginal delivery, and 10 percent a low chance. The false positive rate fell to 0.4 percent (1/263) and the false negative rate also fell to 3 percent (9/263). Thus, Flamm's tool may be preferred, from a diagnostic test perspective, due to an ability to stratify more of the population into high and low probability subgroups with a low false positive rate.

Imaging Modalities

Seven studies^43–49 examined the role of imaging modalities in predicting the outcome of a TOL after prior CD. In these studies a variety of imaging factors were considered, including the two fundamental aspects of labor: passage (pelvic dimensions) and passenger (fetal dimensions).

Four studies^{44, 45, 47, 49} focused primarily on the imaging of the passage using X-ray pelvimetry (XRP). Of these studies, three were retrospective cohorts ^{44, 45, 49} that were given poor-quality ratings because of inadequate control of confounding or effect modifiers, unequal application of measurements, and unidentified patient spectrum composition. The fourth study was a good-quality RCT by Thubisi.⁴⁷ Half of the 288 subjects were assigned to receive an antepartum XRP evaluation; the remaining subjects were allocated to the postpartum XRP evaluation group. Of those in the antepartum group, 84 were considered to have an adequate pelvis and 23 of these delivered vaginally (27.7 percent). All of the patients considered on antepartum XRP to have an inadequate pelvis had an ERCD. Of those in the postpartum XRP group, 41.6 percent (60/144) delivered vaginally. In the postpartum XRP group considered to have an inadequate pelvis based on clinical examination, 60 percent (33/55) had a vaginal delivery, compared with 30 percent (27/89) of those considered to have an adequate pelvis. This study provides strong evidence that XRP is a poor predictor of TOL outcome and might unnecessarily increase CD rates.

Three poor-quality prospective cohort studies^{43, 46, 48} examined the value of a scoring system based on a variety of fetal and maternal pelvic measurements and calculated circumferences (fetal head, fetal abdomen, pelvic inlet, and midpelvis), to predict vaginal delivery. Two^{46, 48} of the three studies that focused on the fetal-pelvic index found that it was significantly associated with vaginal delivery; however, all three studies lacked adequate control for confounders and suffered from verification or workup bias.¹⁸

Summary

• Two validated scoring systems categorized women into groups with likelihoods of vaginal delivery ranging from roughly 45–95 percent.^{36, 40}

• Flamm's tool was able to stratify more of the population into high and low probability subgroups, with a relatively low false-positive rate.³⁶

• By using a prospective cohort design and the largest study population, the best scoring system created a 10-point score based on the presence or absence of five variables commonly available for most patient admissions.³⁶

• An RCT clearly demonstrated the inability of XRP to predict route of delivery reliably.⁴⁷

• Imaging studies that combined the measurements of the pelvis and fetus showed promising results, but were limited by their lack of control for confounding and biases.^{46, 48}

Maternal Complications

Three maternal complications were investigated: major maternal hemorrhage (requiring transfusion or hysterectomy), maternal infection (as manifested by endomyometritis, wound infection, and/or postpartum/puerperal fever), and maternal death (uterine rupture is detailed in question 4). While not all articles addressed each maternal complication, several addressed key aspects of these sequelae.

Two good-quality studies^{5, 20} provided information concerning both transfusion and hysterectomy rates. Rates of maternal hemorrhage requiring transfusion were 1.1 percent in the TOL group versus 1.3 percent for repeat CD in the large population-based study (NS)⁵ and 0.72 percent versus 1.72 percent for the prospective cohort study (p=.0001).²⁰

While several studies provided information concerning hysterectomy, none specifically documented the indication for hysterectomy. Comparisons between TOL and elective CD were reported in three studies.^{5, 20, 30} The best evidence comes from the one large population-based study⁵ that found no difference in hysterectomy rates in TOL (0.2 percent) versus ERCD (0.2 percent). Unlike the two prospective studies reporting this outcome, McMahon attempted to exclude “elective” repeat CDs for medical or obstetric indications such as placenta previa.

The two prospective cohort studies reported higher hysterectomy rates in repeat CD: 0.12 TOL versus 0.27 percent ERCD²⁰ and 0.27 TOL versus 3.2 percent in ERCD.³⁰ These provide weaker evidence because the cesarean group may have included women who had an indication for CD and would not have been candidates for a TOL. In fact, in the latter study, Paul mentions that only 62 of the 157 “elective” repeat CD group were considered to be eligible for TOL. Thus it is possible that the higher rates of hysterectomy could be due to medical or obstetric conditions such as hemorrhage secondary to placenta previa. Hysterectomy rates were reported in only one induction study, reporting 0.2 percent in induced and 0.08 percent in SL patients.²⁸ Overall, there was a trend toward increased risk for hysterectomy in induced labor (increased risk 0.12 percent) and ERCD (increased risk 0–3 percent). These studies did not specify whether hysterectomies were performed for hemorrhage or other indication (cervical cancer, myomatous uterus).

Studies reporting maternal infection rates are limited by lack of explicit definitions or by combining many sources of infection, which make specific clinical insights limited. No study provides data on the risk for spontaneous TOL that is free from medical augmentation. Two studies^{5, 24} defined infection clearly and compared the incidence in TOL and ERCD groups. Both definitions combined puerperal infection and abdominal wound infection. In the larger study,⁵ which defined maternal infection as puerperal fever (temperature >38 degrees C; uterine, urinary, pulmonary, or wound infection; or sepsis) or abdominal wound infection, the rates were 5.3 percent in TOL versus 6.4 percent in ERCD. Subgroup analyses found that women who had a TOL but did not delivery vaginally (e.g. failed TOL), had significantly higher infection rates than women who were able to deliver vaginally (failed TOL 8 percent versus successful TOL 3.5 percent). This finding was reported consistently among prospective cohort studies that performed similar subgroup analyses^{23, 26, 30} (11 to 30 percent increased risk of infection for failed TOL). The other study, a fair-quality prospective cohort,²⁴ reported maternal infection rates (including endomyometritis and wound infection) of 6.79 percent in TOL versus 9.73 percent in ERCD.

Compared with spontaneous onset of labor, there appears to be a trend toward increasing risk of infection when labor is induced (1–4 percent increased risk) and with ERCD (2–3 percent increased risk). However, only one study of induction agents evaluated this outcome, and found zero in the induced group and 5 percent in the SL group.³⁴

Six studies examined maternal death rates. The large population-based study found no maternal deaths in either TOL or ERCD groups totaling 6,138 women.⁵ In five prospective cohort studies involving approximately 19,000 patients, there were two deaths among women having a TOL and two among women having a repeat CD.^{20–23, 27} No maternal deaths were mentioned in any studies of induction of labor (n = 7,525).

Infant Outcomes

APGAR scores. There are insufficient data to compare infant Apgar scores for a TOL versus ERCD. In one fair-quality prospective cohort study,²⁰ more infants born from TOL had 5-minute Apgar less than 7 (1.47 percent versus 0.68 percent, p=.004).²⁰

Infant death. No study has measured infant death directly attributable to a mother's choice of TOL or repeat CD. Two large, population-based studies provide information about whether TOL poses increased risk of infant death compared with ERCD.^{5, 6} Each has important strengths and limitations. One study⁵ (n = 6,138) reported perinatal death rates of 9/1,000 in the TOL group versus 5/1,000 in the repeat CD group for women with one prior CD. The strength of this study was its ability to identify a conceptual cohort of women with one prior low transverse CD who attempted TOL or repeat CD. However, no details were provided on these deaths (e.g., whether infants with lethal anomalies were included), so it is not possible to determine whether these deaths were attributable to labor or cesarean.

A more recent population-based study from Scotland⁶ did exclude all perinatal deaths associated with lethal anomalies and medical conditions; however, they did not do a good job of classifying patients as TOL and ERCD. To ascertain the perinatal death rate attributable to delivery method, the authors excluded all deaths associated with congenital anomalies, antepartum stillbirth (intrauterine fetal death), multiple gestation, and noncephalic presentation. Additionally, they excluded all primary CDs. They divided all remaining deliveries into women with no prior CD who were nulliparous or multiparous, and women with prior CD who delivered by planned repeat CD or TOL. The TOL group was defined as any vaginal delivery or emergent CD regardless of intended delivery route.

There were 20 deaths in 15,515 TOLs for a rate of perinatal death of 12.9/10,000 (95 percent CI, 7.9 to 19.9) versus one in 9,014 repeat CDs for a rate of 1.1/10,000 (95 percent CI, 0.0 to 6.1), and 135 in 137,630 nulliparous women without prior CD for a rate of 9.8 (95 percent CI, 8.3 to 11.6), and 90 in 151,549 multiparous women without prior CD for a rate of 5.9/10,000 (95 percent CI, 4.8 to 7.3). This study is discussed in significant detail in this report because it has not been reviewed in the literature to date.

The authors emphasized that the infant death rate was 11 times higher in women choosing TOL than in those having a CD, corresponding to one additional infant death for every 849 patients. The rate of infant death in women choosing TOL was similar to primiparous women having a vaginal delivery. This would indicate that the woman choosing TOL is not assuming considerable additional risk for her infant in choosing TOL in the second pregnancy. However, the rate of infant death for repeat CD patients appears to be spuriously low. The cesarean group may be low due to misclassification because all emergent CDs and vaginal deliveries were classified as TOL regardless of intended route of delivery. There were 20 perinatal deaths in the TOL group; eight were delivered vaginally and 12 were emergent CDs. If only three of these deaths were misclassified (e.g., women intending elective repeat who required emergent CD), there would not be a statistically significant difference between perinatal death rates in TOL and repeat CD groups. One study examined the rate of emergent CDs in each group.³⁰ They report that two of nine (22 percent) emergent cesareans performed for fetal distress were performed for women who desired repeat cesarean. If this proportion were applied to Smith's emergent cesarean perinatal deaths, three of the 12 would have been expected to occur in the planned repeat cesarean group and 9 in the TOL group. This small change would eliminate the statistically significant difference that was observed. Another potential source of misclassification that would decrease the risk of planned CD compared with TOL is in the antepartum stillbirth data, all of which were excluded.

Even though the authors went to great lengths to consider confounding, there is still substantial detail missing in understanding the context in which these perinatal deaths occurred. For example, the authors were unable to determine the type of prior CD scar (classical, vertical, etc.). To exclude women who might have had classical incisions, they excluded all births that occurred before 40 weeks' gestation, with the thought that women with known prior classical incisions are generally delivered by cesarean before 40 weeks. In confining their sample to those women who delivered at 40 weeks or greater, they might have introduced an additional confounder in that risk of perinatal death increases with higher gestational age, especially 42 weeks and greater. In fact, when they looked at gestation less than 39 weeks versus greater than 39 weeks, they found only three deaths between 37 and 39 weeks, all of which had PGE2 induction of labor. One question that arises is in the group that was greater than 39 weeks' gestation: what proportion of the perinatal deaths were in infants who were 42 weeks' gestation or greater? One of the greatest concerns for women with prior CD is the risk of uterine rupture, and the resulting potential for maternal or fetal morbidity and mortality.

This study did not specifically examine the subset of perinatal deaths attributable to uterine rupture. Uterine rupture was combined with cord compression/prolapse, birth trauma, and asphyxia associated with disproportion in a category called “mechanical” causes. These events are all limited to vaginal delivery; therefore, it is not surprising that the authors found seven perinatal deaths attributed to “mechanical” causes in TOL and none in CD. Additionally, it is not clear how TOL versus planned repeat CD were classified (post-hoc or intention).

Another potential confounder is the use of induction and augmentation agents. The study reports deaths from 1992 to 1997, but does not describe how often induction agents were used, or in what doses, across Scotland during those years. Fifteen percent of their population with prior CD had PGE2 induction of labor. There was no association between PGE2 induction and increased risk of infant death. Although oxytocin was used, the authors were not able to examine whether oxytocin posed any increased risk. Communication with the authors revealed that oxytocin would be used for women with prior CD and premature rupture of membranes, but is not frequently used to augment women for failure to progress during labor.

Importantly, the population-based studies do not describe the likely outcomes of high-quality obstetric care. Even if one accepts that the increased infant death rate in the TOL group is real, the studies do not suggest an answer to the question, “Is there an increased risk of infant death in a properly managed TOL?”

Fifteen studies of induction agents reported infant mortality. Of these, 11 found no deaths in any group studied.^{22, 25, 27, 29, 30, 32, 34–36, 50, 51} In the other four, no consistent pattern emerged favoring spontaneous or induced labor.^{6, 28, 31, 52}

In summary, there appears to be a trend toward increased risk of fetal death for TOL versus ERCD. Although these studies attempted to account for some confounders, their retrospective nature makes it impossible to determine whether the method of delivery is responsible for any increased risk. The validity of the recent publication from Scotland is uncertain because the infant death rate in the CD group appears to be spuriously low, deaths were not directly linked to uterine rupture, some antepartum deaths could have been misclassified, and the TOL group included women who really intended to have an ERCD.

Summary

Increased Risk with Induction

Uterine rupture was reported in 29 of 48 studies of labor induction; however, 15 of these did not report the definition used. Twelve studies reported no cases of symptomatic uterine rupture. Of those studies providing a clear definition of symptomatic uterine rupture and finding any cases of uterine rupture, the lowest rate among the induction groups was 0.35 percent (1 of 289) in a prospective cohort study of oxytocin,³⁰ and the highest was 6.25 percent (1 of 16) in a randomized controlled trial of mifepristone.³⁵ The rates of rupture among women undergoing spontaneous onset of labor in these studies ranged from a low of 0.15 percent in a prospective study of PGE2 gel³⁶ to a high of 0.8 percent in a similar prospective cohort study of oxytocin.²⁸

Figure 8. Uterine Rupture: All Indus tion Method s (more...)

   Figure 8. Uterine Rupture: All Indus tion Method s versus Spon taneous Labor Risk difference, 95% CI

Figure 8

Figure 9. Uterine Rupture: Oxytocin versus Spontaneous (more...)

   Figure 9. Uterine Rupture: Oxytocin versus Spontaneous Labor Risk difference, 95% CI

Figure 9

Figure 10. Uterine Rupture: Prostagiandin s versus (more...)

   Figure 10. Uterine Rupture: Prostagiandin s versus Spontaneous Labor Risk difference, 95% CI

Figure 10

Only one study⁶⁷ compared one induction method versus another. It compared misoprostol to PGE2 (gel or pessary) in a prospective cohort study that did not provide a definition of uterine rupture.⁶⁷ This study found a higher rate of rupture with misoprostol, but the difference was not significant. The largest study of prostaglandin was excluded from analysis due to poor definition of uterine rupture.⁴ Although the precision and accuracy of the results are reduced, the magnitude of the effect showing an increase in the rate of uterine rupture suggests that a real association between PG induction of labor and uterine rupture probably exists.

Predictors of Major Morbidity due to Uterine Rupture

Many have wondered whether there are any factors that can prevent poor neonatal outcome when there are signs of potential rupture. Two fair-quality case series^{55, 56} have studied cases of uterine rupture of the cesarean scar to determine whether any predictive premonitory signs exist. Leung et al. were the first to perform an exploratory analysis to study risk factors for poor neonatal and maternal outcome; particularly FHR and uterine contraction patterns.⁵⁵ They identified 106 cases of symptomatic uterine rupture from 11,179 TOLs in women with prior CD at LA County-USC Women's Hospital, from which they were able to review the records of 99. The scar type was unknown in 99 percent of their population. They categorized cases of uterine rupture based on complete, partial, or no extrusion of the fetus. Combining death, asphyxia, and respiratory distress, they concluded that perinatal morbidity and mortality was significantly greater in cases where the fetus was extruded. However, they report that the six neonates requiring intubation were extubated and discharged from the neonatal intensive care unit (NICU) within 24 hours (range 1–24 hours) and were discharged from the hospital without adverse sequelae. If these six temporary outcomes (e.g., without significant adverse sequelae) are removed, major perinatal morbidity (asphyxia or death) occurred in 7/41 (17 percent) cases of partial or complete extrusion and 4/58 (6.9 percent) cases of nonextruded fetuses (p = 0.113). Of note, four of the fetal deaths occurred in patients who presented with fetal distress and underwent immediate CD, leaving two cases occurring in women undergoing supervised labor (one in the extruded group and one in the nonextruded group). Looking for premonitory signs of uterine rupture, they found that abnormalities of FHR tracing (prolonged deceleration only [defined as FHR less than 90 beats/min that exceeded 1 minute and without return to baseline], prolonged decelerations preceded by late decelerations, prolonged decelerations preceded by severe variables, mild late decelerations only, or fetal distress on admission necessitating CD) occurred in 91/99 cases (91.9 percent) and that all cases of fetal extrusion had prolonged decelerations. Prolonged decelerations occurred in 17/41 (41.5 percent) patients with extrusion and 15/58 (25.9 percent) without. In studying patients with prolonged deceleration further, they found that no patient who had prolonged deceleration only as their sign had significant clinical morbidity when delivery occurred within 17 minutes of the onset of deceleration. If the three cases of temporary neonatal intubation were removed, one case of neonatal asphyxia and no deaths in the prolonged bradycardia group would remain. Although the small numbers make the data unstable, it is intriguing that the one case of asphyxia occurred when there was 32 minutes between the onset of bradycardia and delivery, compared with 22 minutes and less in the group with intubation or no complications. Thus it is unknown what neonatal outcomes would arise between 22 and 32 minutes from bradycardia.

Leung et al. have done a superb job of exploring the details of their cases of uterine rupture; however, they are limited by the constraints of case series data. Data from a control group are important for understanding details about the association between fetal bradycardia and poor infant outcome. Decelerations are not rare; in fact, only 1.4 percent of all deliveries do not have FHR decelerations.⁶⁹ Prolonged decelerations, especially given Leung's definition, are rare, occurring in 7.9–12.5 percent of patients receiving epidurals.⁷⁰ Causes of prolonged decelerations include cervical examination; rapid decent in the second stage of labor; maternal hypotension due to positioning, epidural, or other; maternal hypoglycemia; reactive hypothermia such as with a cold amnioinfusion; prolonged cord compression (oligohydramnios); tetanic uterine contractions; maternal seizures, and cord prolapse, in addition to uterine rupture. Because fetal bradycardia is not specific to uterine rupture, the presence of a control group would allow some insight into associations with uterine rupture versus these other causes. Additionally, it is important to know details about the context of decision-making, in order to know what portion of time delays are preventable (e.g., substantial time between decision to go to cesarean and actual time for cesarean).

A second and more recent case series found no relation between time from FHR deceleration and infant outcome.⁵⁶ All medical records in a single-institution hospital were examined to identify cases of “complete cesarean scar disruption,” defined as uterine scar separation that extended through visceral serosa. As above, the study was conducted in a tertiary care hospital with in-house anesthesia and obstetrics. The authors report on 23 cases of uterine rupture of a cesarean scar, six with partial or complete expulsion of the fetus. Fetal heart rate abnormalities—which included tachycardia and late, variable, or prolonged (not defined) decelerations—were the initial sign of uterine rupture in 87 percent of cases (four had pain, one vaginal bleeding, and one hematuria). Prolonged deceleration was the first sign of uterine rupture in 6/6 (100 percent) of the extruded patients versus 8/17 (47 percent) without extrusion. There was one perinatal death that occurred in the non-extruded group (late decelerations more than 25 minutes before delivery, failed vacuum extraction, then cesarean), and three cases of impaired motor development diagnosed as hypoxic-ischemic encephalopathy, occurring in the extruded group; delivery occurred 15,16, and 23 minutes from onset of prolonged deceleration. When they looked at metabolic acidosis (their primary outcome, defined as umbilical artery pH less than 7.0 with base deficit greater than 12mMol/L), they found a non-significant trend towards less time between first sign to delivery (18 versus 24 minutes) and decision to delivery (13 versus 17 minutes) in the group with metabolic acidosis compared with those without acidosis (p = 0.11). In this case, the greater time delays in the group without metabolic acidosis could reflect less concern by the physician and thus a slower overall movement, rather than programmatic delays.

In summary, the literature on uterine rupture suffers from inconsistent use of terms and ambiguous definitions. Additionally, because uterine rupture of the cesarean scar is often diagnosed at cesarean performed for fetal tracing abnormalities, there is diagnostic review bias. Studies conducted thus far to examine the relationship between duration of FHR disturbance particularly prolonged bradycardia and adverse perinatal outcome, have had conflicting results. It is important to further examine the relationship between fetal tracing disturbances (e.g., prolonged fetal bradycardia) and uterine rupture. This can only be done by comparing instances of a particular fetal tracing disturbance in women undergoing a TOL and noting how many times it is truly associated with uterine rupture (true positive) and how many times it is not (e.g., false positive).

Summary

• The use of terms among studies is inconsistent.

• Definitions of terms among studies are ambiguous.

• There is not a significant difference in asymptomatic uterine rupture rates in TOL versus ERCD.

• Symptomatic uterine rupture is significantly more common in TOL versus ERCD, with an increased risk of 2.7/1000

• Based on the frequency and severity of symptomatic uterine rupture, the risk of perinatal death due to a rupture of a uterine scar is 1.5/10,000 and the risk of hysterectomy is 4.8/10,000. These rates of serious complications such as perinatal death, are probably more precise than overall risks from studies measuring death directly.

• The definition of uterine rupture as an outcome is confounded by a definition that includes the potential predictor of FHR tracing abnormality.

• Measurement of frequency of occurrence, predictors for what population is at greatest risk, and predictors for poor outcomes are difficult, because of the lack of standard case definition.

Summary

• There were no studies of health status or health-related quality of life for VBAC or repeat CD patients.

Summary

• Studies of patient satisfaction largely consisted of patient's own provider obtaining information about satisfaction, introducing the possibility for measurement bias.

• Only two cross-sectional studies used methods other than the patient's own provider to obtain satisfaction information.

• No study measured satisfaction for the three types of delivery outcomes that could be experienced by women with prior CDs (VBAC, TOL followed by CD, or ERCD), which leaves room for much needed research.

Summary

• Based on the economic evaluation with the best quality score, when the probability of vaginal delivery is 76 percent or greater, TOL is more cost-effective and provides higher quality of life.

• Based on the economic evaluation with the best quality score⁸⁷ and assuming costs per QALY of $50,000 as cost-effective, the more cost-effective of TOL and ERCD depends on the probability of successful VBAC after TOL.

• Further evaluation is needed of the sensitivity of the probability cut point of 76 percent to other potential predictor variables.

Health Care Resources

Past Obstetric Factors

While over 50 studies have investigated the influence of clinical history and past obstetric factors on the outcome of TOL after prior CD, only five were of fair-to-good quality. This relatively small percentage of quality studies did not provide any information for the individual factors of gravidity, parity, and previous cervical dilation.

Table 10. Past Indicators of VBAC Delivery

Table 10. Past Indicators of VBAC Delivery

Factor	Author (year)	Adjusted OR for VBAC	95 percent CI p-value
Prior VD	McNally 1999107	27.78	3.85–200
Order of prior VD
Before prior CD	Flamm 199736	1.53	1.12–2.10
	Weinstein 199642	1.8	1.1–3.1
After prior CD	Flamm 199736	3.39	2.25–5.11
	Macones 200138	7.69	3.23–20
After vs. Before prior CD	Caughey 1998112	3.48	1.9–6.1
Before & After prior CD	Flamm 199736	9.11	2.18–38.04
Number of prior CD	Pickhardt 199239	0.43	p<0.05
Prior CD Indication
Nonrecurrent vs. Recurrent	Flamm 199736	1.93	1.58–2.35
Recur vs. Nonrecurrent	Weinstein 199642	0.8	0.3–2.0
Breech vs. Recurrent	Weinstein 199642	1.9	1.0–3.6
Fetal Distress vs. Recurrent	Weinstein 199642	1.05	0.4–2.6

Bold=significant; NR=not reported; NS=not significant

When considering the issue of prior CD, the two most investigated factors include the number of prior CDs and prior CD indication. Of the 22 studies looking at the number of prior CDs, only one was rated as being fair in quality.³⁹ Consistent with the overall literature, Pickhardt³⁹ demonstrated that the probability of VBAC significantly decreased as the number of prior CDs increased (adjusted OR 0.43; p < 0.05). By controlling for a great number of potential confounders in his analysis, Pickhardt established this factor as a true independent predictor of TOL outcome. Also consistent with the overall VBAC literature were the findings of the two^{36, 42} of 61 studies given a fair rating regarding prior CD indication. While Flamm³⁶ demonstrated that those with a nonrecurrent indication compared with those with a recurrent prior CD indication (CPD or failure to progress), had a significantly higher VBAC rate (adjusted OR 1.93; 95 percent CI, 1.58 to 2.35), Weinstein⁴² showed similar, yet nonsignificant findings. Weinstein also found that although nonsignificant, those with a prior CD indication of breech presentation or fetal distress had a greater chance of VBAC compared with those with a recurrent indication (adjusted OR 1.9; 95 percent CI, 1.0 to 3.6 and adjusted OR 1.05; 95 percent CI, 0.4 to 2.6, respectively). As reported by previous studies, those with a prior CD indication of breech presentation had the highest relative likelihood of VBAC.

Current Obstetric Factors

We found no fair or good studies addressing the factors of multiple gestations, cervical dilation rate, SL, induced labor, oxytocin use, maternal height, maternal weight, and maternal weight gain.

Three case series provide the only data regarding the association between external cephalic version (ECV) and VBAC.^{104, 105, 108} Rates for VBAC after ECV attempts ranged from 65.8 to 100 percent. By comparing ECV attempts in those with prior CD to those without prior CD, Flamm¹⁰⁵ showed that those with prior CD were significantly more likely to be successfully verted (82 percent and 61 percent, respectively, p = 0.02). Although the overall VBAC rate in these three studies ranged from 50 to 54.5 percent, de Meeus¹⁰⁴ showed that of those who had a successful version, the VBAC rate was actually higher (76 percent). Another finding of interest came from the Schacter¹⁰⁸ study, which found that those delivering within a week of ECV had a significantly lower VBAC rate compared with those who delivered more than a week after ECV (0 percent [0/4] and 86 percent [6/7], respectively).

Table 11. Current Indicators of VBAC Delivery

Table 11. Current Indicators of VBAC Delivery

Factor	Author (year)	Adjusted OR for VBAC	95% C p-value
Gestational Age	Pickhardt 199239	0.81	p < 0.05
	Zelop 2001110	0.67 (>40wks GA, spontaneous)	0.56–0.83
	Zelop 2001110	0.67 (>40wks GA, induced)	0.45–0.91
Birth weight	Weinstein 199642	0.95 (>4000g)	0.17–5
	Zelop 2001111	0.59 (>4000g)	0.45–0.77
Cervical Dilation	Flamm 199736	2.16 (>4cm)	1.66–2.82
	Macones 200138	1.87	1.14–3.23
	Pickhardt 199239	1.62	p < 0.05
	Stronge 1996109	NR	NS
Effacement	Flamm 199736	2.72 (>75%) – referent <25 percent	2.00–3.71
	Flamm 199736	1.79 (25–75%) – referent <25 percent	1.31–2.44
	McNally 1999107	5.0 (100%)	1.28–19.23
Station	Stronge 1996109	12.3	4.6–33.3
Bishop score	Weinstein 199642	6.0 (score >4)	3.5–10.4
Augmentation	Macones 200138	0.47	0.25–0.88
	Stronge 1996109	NR	NS
Epidural use	McNally 1999107	0.26	0.06–1.12

Bold=significant; NR=not reported; NS=not significant

The effects of various medications on TOL outcome have been one of the more heavily investigated areas of VBAC literature. No fair-to-good-quality studies provided information regarding labor induction or oxytocin use (in general); however, of 21 studies that provided information regarding the factor of labor augmentation, there were two fair-quality studies.^{38, 109} Although Macones³⁸ demonstrated that those with labor augmentation were significantly less likely to have VBAC compared with those without augmentation (adjusted OR, 0.47; 95 percent CI, 0.25 to 0.88), Stronge¹⁰⁹ found no significant association between labor augmentation and TOL outcome. Once again, one could speculate that this difference in results could be due to a lack of power in Stronge's study to find an association or perhaps due to a differential level of confounding adjustment. Of the 16 studies to investigate the influence of epidural use on the outcome of TOL, only one was of fair quality. Although nonsignificant, McNally¹⁰⁷ demonstrated that those with the use of an epidural tended to have a lower likelihood of VBAC compared with those who did not use an epidural.

NonClinical Factors

Although medical decisions are often based on clinical factors alone, it is important to remember that nonclinical factors might also play an important role in VBAC. For example, McMahon⁵ found that those who attended prenatal classes were significantly less likely to fail a TOL compared with those who did not attend (crude OR, 0.8; 95 percent CI, 0.6 to 0.9). In addition to this, Fraser¹⁰⁶ conducted a fair-quality RCT comparing the effect of either a verbal-based (individualized discussion program) or a document-based (pamphlet) prenatal program for those attempting a TOL after prior CD. Although statistically nonsignificant, the results showed that those in the verbal treatment arm had a higher rate of VBAC compared with those in the document treatment arm (53 percent and 49 percent, respectively; RR, 1.1; 95 percent CI, 1.0 to 1.2). This review investigated the influence of three nonclinical factors (i.e., insurance, physician characteristics, and hospital characteristics) on the outcome of a TOL after prior CD.

While a number of studies in the VBAC literature provided information regarding the nonclinical factors of insurance status, physician characteristics, and hospital characteristics, none of them were of fair-to-good quality. The majority failed to adjust for confounding (e.g., Socol¹¹³, McMahon⁵); those that did provide adjusted ORs (e.g., Goldman,¹¹⁴ King,¹¹⁵ Stafford¹¹⁶) did so using database information that limited them to the comparison between those with VBAC and those with CD, which included those with either an ERCD or a failed TOL.

Summary

• The vast majority of studies looking at individual factors that influence the route of delivery were of poor quality due to inadequate control for confounding factors.

• The factors that were significantly associated with an increased likelihood of vaginal delivery (i.e., successful TOL) were: maternal age less than 40 years,³⁶ PRIOR VD (particularly vaginal delivery after cesarean),^{36, 38, 42, 107} a nonrecurrent indication for the prior CD,³⁶ and favorable cervical factors.^{36, 38, 39, 42, 107, 109}

• The factors that were significantly associated with a decreased likelihood of vaginal delivery (i.e., failed TOL) were: an increasing number of prior CDs,³⁹ gestational age greater than 40 weeks^{39, 111} birth weight greater than 4000 g,¹¹¹ and augmentation of labor.³⁸

Factors Relating to Patient's Birth Choice and Reasons for Choice

Before patient preferences were assessed, the proportion of women who actually had a choice was determined for each study. The proportion of eligible women (minimal requirement: low-transverse scar, singleton fetus, and no other contraindications) choosing to attempt a TOL ranged from 22.6 to 90 percent in the six fair-to-good-quality studies that were clear about the inclusion/exclusion criteria.^{24, 57, 106, 119, 121, 128} As might be expected, the two studies conducted in the early 1980s^{24, 57} had much lower attempt rates (22.6 to 31.5 percent) compared with the other four studies, which were conducted between 1989 and 2001 (attempt rates 42 to 90 percent).^{106, 119, 121, 128}

In total, 1,083 of 2733 eligible women in six studies chose TOL (sample weighted average of 39.6 percent).^{24, 57, 106, 119, 121, 128} The VBAC rate for eligible women choosing TOL ranged from 56.5 to 84.5 percent. In total, 778 of the 1,083 eligible women had a VBAC (sample weighted average of 71.8 percent).

The heterogeneity of the inclusion criteria (when they were assessed) might have contributed to variation in the proportion choosing a TOL. In three studies the women were pregnant and had a history of a prior CD when preference was assessed.^{24, 119, 122} In three studies the women were assessed within days of delivery.^{84, 106, 118} In one study the assessment was within 1 month of delivery,¹²⁴ and in one study the women were assessed several months after delivery.⁵⁷ Finally, in one study the women were interviewed both when they were pregnant and postpartum.¹²⁰

Several factors (race, prior VD, social motives, safety, future childbearing plans) appeared to influence choice of delivery. The proportion of nonwhite patients ranged from 2.4 to 47 percent in the four fair-quality studies that reported race.^{118, 120–122} Only one prospective cohort study of good quality examined the effect of race on preference.¹²⁰ In this study 23/43 (53.5 percent) nonwhite patients attempted a TOL and 42/50 (84 percent) of white patients attempted a TOL. Forty-seven percent of the nonwhite patients were black, 28 percent were Latino, and 21 percent were Asian. All women in this study were middle-class and working class women. Although the white patients were more educated than the nonwhite patients, all other socioeconomic status indicators were similar. Several results in this study suggested that the minority patients had less opportunities to gain medical information about delivery options than white patients. Fewer minority patients attended childbirth courses (43 percent) during their first pregnancy when compared with white patients (81 percent) (p < .0001). Compared with white patients, minority patients were less likely to have been told by their former providers after their prior CD that VBAC was possible (p < .003). Even though minority patients received less medical information and encouragement for a TOL, more patients (39 percent) identified the provider as an important influence in their decision, compared with 19 percent of white patients (p < .02).

In addition to informational differences between the races, underlying cultural ideologies might account for the different approaches to delivery.¹²⁰ From structured interviews, these investigators reported that ethnic minority women viewed labor as a painful necessary evil that does not relate to one's intrinsic worth. Forty-six percent of minority patients did not want to experience labor again compared with 22 percent of white patients. If a woman could become a mother through a less painful, less risky manner, e.g., with an ERCD, no one look downed on them. By contrast, these same investigators described the view of labor by white patients as a challenge to be overcome to gain full status as mothers. White women viewed vaginal birth as a “once-in-a-lifetime experience not to be missed.”

Two of 11 studies examined prior VD as a predictor for a TOL preference.^{24, 123} In both studies, patients who had delivered at least one baby vaginally were more likely to choose TOL. A greater proportion of the women choosing TOL had a history of vaginal delivery either before or after their CD (18/53, 40.0 percent) when compared with women who chose ERCD (only 5/46, 10.9 percent had prior VDs) (p = 0.007).¹²³ Possibly, women who have already succeeded with a vaginal delivery have a stronger self-efficacy or belief that by doing a TOL they will indeed deliver the baby vaginally. One cross-sectional study that examined state anxiety reported that women choosing TOL had lower state anxiety and felt better prepared than women choosing ERCD.¹²² Four of 11 studies cited fear of labor or fear of failure as a strong reasons for choosing ERCD.^{57, 118, 123, 124} These patients felt that a TOL would lead to a difficult labor, failure to deliver vaginally, and, in the end, another CD.⁵⁷

Social motives (ability to care for children at home, convenience) appeared more often in these studies as the primary reason for selecting TOL or ERCD than careful weighing of health risks for mother or baby. Six of the seven studies that reported patients' reasons for choosing TOL cited “easier recovery” as a strong reason.^{84, 118, 120, 122–124} Women in these studies already had children at home who needed care, so a shorter delivery was very desirable. Five of the six studies reported that the women wanted to experience a vaginal birth.^{84, 118, 120, 122, 124} Structured interviews with women before delivery and 2 months after delivery showed that the women also chose TOL so their husbands could be more involved.^{128, 129} Finally, two of 10 studies cited convenience as a primary reason for ERCD.^{57, 118} A scheduled delivery allows mother and provider to set a date that coordinates well with work and allows time to plan for childcare.

Safety for the mother and/or baby was cited as an important reason in only four of the 11 studies reporting reasons for deliveries.^{84, 118, 122, 124} In a cross-sectional survey of women who had just delivered healthy babies either by ERCD or VBAC, 18/21 women who chose and delivered by VBAC felt that vaginal delivery was safest for the mother compared with 7/11 women who chose and delivered by ERCD.¹²⁴ In this same group of mothers who chose and delivered by VBAC, 10/21 felt vaginal delivery was safest for the infant also, compared with 2/11 who chose and delivered by ERCD. Since this study only recruited women with healthy babies, the results are potentially biased in that the patients tended to believe the method was safe because the outcome was good. Another study using structured interviews showed that the women did not know actual probabilities or complication rates when they made their decisions.¹²⁹ It was unclear if the provider had told them the probabilities and they did not recall them or place importance on them, or if the patients were never informed of the actual probabilities.

Only one good-quality RCT¹⁰⁶ and two fair-quality prospective cohort studies^{24, 120} examined the effect of future childbearing plans on the birthing preference. In the RCT, 23 percent of women with a low motivation for a TOL desired to have a ligation sterilization compared with the 13 percent of women with a high motivation for TOL.¹⁰⁶ In one prospective cohort study,¹²⁸ 22/56 (39.3 percent) women having an ERCD had their tubes tied after delivery, compared with 4/44 (9.1 percent) of women delivering vaginally. Similarly, more women having an ERCD, 245/547 (44.8 percent) requested a ligation sterilization, compared with 18/101 (17.8 percent) of women experiencing VBAC, and 14/61 (23.0 percent) choosing TOL but having a CD.²⁴

Education, Hospital, and Physician Influence on Patient Delivery Choices

The confidence a woman has to succeed at TOL might also be related to how knowledgeable she is about VBAC, particularly before she becomes pregnant or early in her pregnancy. Only three of the 11 studies with valid results described an education process for women with prior CD.^{106, 120, 121} The best-quality study, a good-quality RCT,¹⁰⁶ reported that overall there was no difference in the proportion of eligible women attempting a TOL when given a pamphlet at 21 weeks' gestation versus an individualized VBAC education and support program started at 21 weeks' gestation. However, when the subgroup of patients with very low motivation for TOL was educated and given support, more patients, 28/86 (32.6 percent) chose TOL than the very low motivated patients who received pamphlets (18/93, 19.4 percent) (RR, 1.7; 95 percent CI, 1.0 to 2.8, p = 0.043). The investigators also commented that it was possible that the intervention was launched too late to influence the patient's choices. Indeed, 28 to 49 percent of patients in four other studies had decided to attempt a TOL before the pregnancy began.^{84, 118, 123, 124} Another 34 to 40 percent of patients decided to attempt a TOL before the midpoint of their pregnancy.^{118, 124} The results of these studies suggest that education should be started shortly after the first CD, perhaps at the first postnatal visit.¹²³ In contrast, only 0 to 15 percent of the women in two studies had decided to have a ERCD before their pregnancy began, but 25 to 42 percent had selected it by the middle of the pregnancy.^{118, 124}

The likelihood of VBAC counseling also appears related to the overall CD rate of the hospital the patient chooses for delivery. One fair-quality retrospective cohort study of 51 California hospitals reported that hospitals with higher overall CD rates had higher rates of ERCDs without documented evidence of counseling regarding TOL.¹²¹ In this study, 1,662 birth records were randomly selected from 11 “high CD” hospitals (average CD rate of 30 percent), from 32 “intermediate CD” hospitals (average CD rate of 21 percent), and from eight “low CD” hospitals (average CD rate of 15 percent). Of women eligible for TOL who chose ERCD, 21 percent of women at the “high CD” hospitals had no documented proof of counseling, compared with 15 percent of “intermediate CD” hospitals and 0.3 percent of “low CD” hospitals (p < 0.01 for the three proportions). Another 36 percent of women at “high CD” hospitals were counseled but refused TOL, compared with 29 percent at “intermediate CD” hospitals and 10 percent of women from “low CD” hospitals (p < 0.01 for three proportions). The study further reported that once a patient had been counseled and consented to a TOL, she had a similar chance of a vaginal delivery regardless of the underlying hospital CD rate.

The patient's exposure to VBAC education appears related not only to the hospital she chooses for delivery but also to her own specific physician. The specific wording the provider uses in discussing TOL with patients is difficult to document and might reflect the provider's underlying preferences. In one retrospective cohort study of the general birthing population (not focused on patients with prior CD) for 11 physicians, the variances for CD rates were not explained by patient obstetric risk factors, socio-economic status, service status, or physician's experience, suggesting that the physician's own practice style might influence route of delivery.¹³⁰ In a cross-sectional study of 19 public hospitals in Italy, obstetricians would chose TOL if they worked at a large hospital (delivered more than 1,000 babies/year) (p < 0.01), and if they worked at a hospital with a CD rate of less than 25 percent (p < 0.001).¹³¹

The education and support for TOL a patient perceives from her physician might also be related to her ethnicity. In one fair prospective cohort study, 60 percent of nonwhite patients were aware of a VBAC option before the pregnancy, compared with 86 percent of white patients (p < 0.003).¹²⁰ Seventy-two percent of white patients felt they received “some to much” information and encouragement by their provider on attempting a TOL, compared with 50 percent for nonwhite patients (p < 0. 005). Although white patients perceived that they received sufficient information, a lower proportion of white patients placed great value on their physician's information than nonwhite patients. Thirty-nine percent of nonwhite patients in one prospective cohort study felt the doctor was an important influence, compared with 19 percent of white patients (p < 0.02).¹²⁰

Summary

• Patient preferences for birth choice are unclear because of the heterogeneity of the 11 included studies.

• Several factors appear related to choice for TOL (white race; prior VD; lower levels of anxiety during the pregnancy).

• Lack of medical information along with cultural ideologies might account for minority women being less likely to attempt a TOL when compared with white women.

• A woman's choice for delivery was often based on social motives (e.g., easier recovery, so she can care for baby and children at home).

• Only four of 11 studies cited safety for mother or baby as important reasons for delivery choice.

• It remains unclear if VBAC education increases the proportion of women who choose TOL. Future studies of education should include education before next pregnancy, perhaps at the postnatal visit of patients with first CD. Future work should also insure that all patients regardless of race receive the same information.

Legal or Legislative Characteristics

King and Lahiri¹¹⁵ considered a variety of medical and socioeconomic predictors of rates of VBAC including two variables related to professional liability. These two variables were annual average paid loss (for years 1985-1989) of the hospital due to malpractice claims settlements divided by patient days and the mature-claims-made rate for OB/GYNs in the county of the hospital. A multiple logistic model to predict the probability of VBAC was developed. This model adjusted for a variety of patient demographic and socioeconomic characteristics and for hospital characteristics. The authors fit models with and without data from New York City to determine whether the influence of a characteristic on the results was due largely to New York City. Hospital-paid loss due to practice claims was statistically significant when New York City patients were excluded (OR, 0.96; 95 percent CI, 0.95 to 0.98) but not when New York City patients were included (OR, 1.01; 95 percent CI, 0.99 to 1.03). The physician's premium was statistically significant with the inclusion of hospitals in New York City (OR, 0.98; 95 percent CI, 0.97 to 0.99 for risk of a $5,000 increase in annual premiums) but not when New York City hospitals were excluded (OR, 1.01; 95 percent CI, 1.00 to 1.08). No summary statistics are provided to facilitate interpretation of these ORs and inclusion of interaction terms for New York City would have been more useful. Whether these ORs are statistically significant, the magnitude of the OR is small, indicating relatively little impact on rates of VBAC. While the professional liability variables are statistically significant, since the odds ratios are close to 1.0 they may not be very meaningful.

These two studies provide little evidence of the impact of legal or legislative components on rates of VBAC. For the paper by King and Lahiri,¹¹⁵ the effect of hospital paid loss due to malpractice claims settlements and physician's malpractice premiums were relatively small (OR very close to 1.0). Changes observed in VBAC rates in Studnicki et al.¹³² occurred only in some risk strata. There are not studies regarding the impact of the current malpractice crisis on availability of obstetric providers and impact on a patient's options. Thus additional research needs to be conducted to determine the influence of legal and legislative factors on changing provider behavior relative to type of delivery.

Guidelines

Bickell et al.¹³⁴ selected a random sample of 45 hospitals in New York to receive a program of peer review and audits of 100 cases of labor and delivery with feedback These hospitals were compared with the remaining 120 hospitals in the state to determine differences in VBAC and repeat CD rates. While there was a significant difference in the overall CD rate, there were no significant differences in rates of VBAC or repeat CD, when comparing the year before audits began (1988) with the year after the audits and feedback were completed (1993) There were no differences in baseline characteristics reported and no adjustment was made for potential confounders.

There was one retrospective cohort study rated fair. Santerre,¹³⁶ using data from a group of 55 hospitals in Massachusetts, performed a regression analysis on VBAC rates over 9 years (1985-1993) during which time the ACOG guidelines were published (in 1988). Using a model that adjusted for potential confounding variables including some baseline risk factors (e.g., low birth weight, race, and source of payment), the model predicted a “permanent” 5.6 percent increase in VBAC rate attributable to the guidelines.

Lomas et al. ¹³⁵ also compared average monthly change in rates of repeat CD in Ontario for 6 years before and two years after publication of guidelines recommending reductions in the rates of CD. The guidelines were a Canadian national consensus statement similar to the National Institutes of Health 1980 consensus conference in the US. The rates of repeat CD decreased at a higher rate after the guidelines than before. As these authors did not fully describe the other variables included in their regression model, this study was rated fair.

The study¹³³ that provides the best evidence suggests that use of opinion leaders provides a greater likelihood of changing practice compared with audit and feedback. A recent conference summary¹⁴² echoed this view when it concluded that involvement of opinion leaders is an important step in achieving local buy-in for guidelines. Another study¹³⁴ of peer review and audit failed to demonstrate a significant change in the rates of either VBAC or RCD. The other two studies^{135, 136} suggested that publications of national guidelines do impact practice although perhaps not to the degree expected.

Provider Characteristics

Hospital Characteristics

Gregory et al.¹⁶⁴ compared VBAC rates across hospital settings in California in a study that was rated good. Rates of VBAC (adjusted for baseline and medical characteristics of mother and fetus) were 14 percent in private nonteaching hospitals, 57 percent in public hospitals, 60 percent in private teaching hospitals, and 41 percent in health maintenance organizations (HMOs). When compared with private, nonteaching hospitals, the repeat CD rates in other types of hospitals was statistically significantly different (p < 0.001). The adjusted repeat CD rates were 85.7 percent in private, non-teaching hospitals (the reference group), 43.0 percent in public hospitals, 40.0 percent in private teaching hospitals and 59.0 percent in HMOs.

McMahon et al.⁵ compared rates of TOL and VBAC with type of hospital in Nova Scotia. Compared with tertiary care centers, the ORs for TOL rate were 0.5 (95 percent CI: 0.5 to 0.6) for regional hospitals and 0.4 (0.3 to 0.5) for community hospitals. The ORs for successful TOL were 0.7 (0.6 to 0.8 and 0.5 to 0.9, respectively) for both regional and community hospitals, compared with the tertiary care centers.

Stafford¹¹⁶ reported on relationships between several hospital characteristics and rates of VBAC. The study was rated good and represented all relevant discharges in California in 1986. Across hospital ownership types (compared with proprietary hospitals), the adjusted ORs for VBAC were (1.4; 95 percent CI, 1.2 to 1.6) for private nonprofit hospitals, 3.9 (3.3 to 4.6) for Kaiser Permanente hospitals with Kaiser payment, 2.6 (1.4 to 4.6) for Kaiser Permanente hospitals without Kaiser payment, 2.5 (2.1 to 2.9) for county hospitals with indigent payment, 2.7 (2.1 to 3.5) for county hospitals without indigent payment, and 3.7 (3.0 to 4.6) for the University of California hospitals. Compared to nonteaching hospitals, the adjusted ORs for VBAC were 0.7 (0.6 to 0.8), 0.9 (0.8 to 1.0), and 1.7 (1.5 to 1.9) for nonmedical-school-affiliated teaching hospitals, medical-school-affiliated hospitals, and Council of Teaching Hospitals member hospitals, respectively. Compared with a hospital without an NICU), the adjusted OR for VBAC for a hospital with an NICU was 0.9 (0.8 to 1.0). Across four categories of annual numbers of births, rates of VBAC increased with increasing numbers of annual births.

King and Lahiri¹¹⁵ assessed the impact of various hospital factors on the VBAC rates in New York hospitals in a study rated good. Compared with voluntary hospital ownership, church hospitals had a higher OR (1.13; 95 percent CI, 1.01 to 1.26) of VBAC compared with ERCD. The odds ratio was not significantly different from 1 (1.07; 95 percent CI, 0.95 to 1.21) if New York City hospitals were excluded. Government hospitals had a lower OR (0.77; 95 percent CI, 0.63 to 0.94) and this association did not change if New York City hospitals were excluded. Odds ratios increased with increasing levels of care from I (reference) to II (1.30, 95% CI: 1.18 to 1.44) to III (1.55; 95 percent CI, 1.34 to 1.81). The OR for teaching hospitals was 1.11 (0.99 to 1.24) compared with nonteaching hospitals although not significantly greater unless New York City hospitals were excluded (OR 1.36; 85 percent CI, 1.21 to 1.54).

Santerre¹³⁶ evaluated various predictors for rates of VBAC in a panel of 55 hospitals in Massachusetts in a study rated fair. The authors were specifically interested in ACOG guidelines but they also controlled for other factors, including hospital characteristics. Their model estimated lower VBAC rates at hospitals with a higher proportion of low birth weight babies, hospitals with a higher percentage of Hispanic babies, and nonteaching hospitals. Volume of births, presence of neonatal ICU, ownership status, and urban location did not predict VBAC rate in their model.

Shiono et al.¹⁶³ surveyed a random sample of US hospitals in a study rated fair. They reported rates of TOL adjusted for size of the delivery service (the stratification variable). Adjusted TOL rates were 12.5 percent and 6.5 percent in hospitals with and without NICUs, respectively. Rates for TOLs were 14.6 percent and 6.6 percent in hospitals with and without OB residency, respectively. Rates of TOLs and VBAC increased with increasing size of delivery service, but rates of successful TOLs were highest in hospitals with the smallest (less than 500) and largest (5,000 or more) number of annual deliveries.

The three descriptive studies^{29, 157, 162} of hospital characteristics were all rated fair. These evaluated VBAC in small rural hospitals. Raynor²⁹ reported on the VBAC rate in a small rural hospital in North Carolina. The rate of TOL in 67 eligible patients was 76 percent and the rate of VBAC among these was 61 percent. Two uterine ruptures were reported in this study but neither was related to labor. Schimmel et al.¹⁶² reported on a nurse-midwife service in a rural county in California. Among 37 patients, the VBAC rate was 87 percent and no uterine ruptures were reported. While these studies are small, they provide some evidence of the success of VBAC in rural settings. The third descriptive study was conducted by Walton et al.¹⁵⁷ at an isolated US military hospital in Japan. Of 62 patients, 79 percent agreed to a TOL but 14 failed to meet guidelines for VBAC. Of the remaining 32, 88 percent achieved a VBAC. No uterine ruptures were reported. These reports, while limited, suggest that VBAC might be safely attempted in small rural hospitals. However, the effects of an adverse outcome of a TOL in a small rural setting have yet to be defined.

The comparative studies suggest there are some differences among types of hospital ownership with respect to rates of VBAC. However, categorization of hospital types varied across studies makes comparisons across studies difficult. Gregory et al.⁶¹ reported higher rates of VBAC in public hospitals and private, non-teaching hospitals, and lower rates in private, non-teaching hospitals. Stafford¹¹⁶ found statistically significantly higher rates of VBAC in Kaiser-affiliated hospitals, county hospitals, and University of California hospitals, compared with proprietary and private, nonprofit hospitals. King¹¹⁵ found that, compared with voluntary ownership, rates of VBAC were statistically significantly higher in church-affiliated hospitals and lower in government-affiliated hospitals. McMahon et al.⁵ found statistically significantly lower ORs for VBAC in regional and community hospitals, compared with tertiary medical centers. Santerre¹³⁶ found no statistically significant association of type of hospital ownership with VBAC rates. Thus, additional research is required to clarify this potential association.

With respect to hospitals with teaching programs, Gregory et al.⁶¹ found private teaching hospitals had statistically significantly higher rates of VBAC than private non-teaching hospitals. King and Lahiri¹¹⁵ estimated an statistically non-significant OR of 1.11 comparing VBAC and ERCD in teaching versus non-teaching hospitals. Stafford¹¹⁶ found the highest OR for VBAC versus ERCD at hospitals that were members of the Council of Teaching Hospitals but ORs for other teaching hospitals (whether or not they were affiliated with medical schools) were lower than for non-teaching hospitals. Santerre¹³⁶ found a statistically significantly lower VBAC rate among non-teaching hospitals than teaching hospitals. Shiono et al.¹⁶³ estimated that hospitals with OB residency programs had statistically significantly different rates VBAC rates about twice as high as those that did not. Thus, as with ownership above, some studies suggest that teaching hospitals have higher rates of VBAC than non-teaching hospitals, but the association does not hold across all categorizations of teaching versus nonteaching.

With respect to the association of an NICU with rates of VBAC, Shiono et al.¹⁶³ estimated VBAC rates were about twice as high in hospitals with an NICU compared with hospitals without an NICU. Stafford¹¹⁶ found an OR of 0.9 comparing hospitals with an NICU with those without (for VBAC versus ERCD). Santerre¹³⁶ found no significant association of the presence of an NICU with VBAC rate. Thus if there is an association of the presence of an NICU with VBAC rate, this association is not consistent across studies.

Across several hospital characteristics, there are no consistent associations with rate of VBAC. This might reflect lack of consistent definitions of categories across studies (e.g., types of hospital ownership), changes in these categorizations over time, a variation in the potential confounding variables that were controlled for in each study, or other factors.

As discussed in the patient preferences section, the decision between a TOL and ERCD is generally made prior to arrival at the hospital for delivery. Thus some hospital characteristics are likely to be confounded with other health care system characteristics (or patient or clinical status characteristics). In particular, providers affiliated with a particular type of hospital might exert much more influence on the decision for TOL or ERCD than the hospital itself. To the extent that a specific type of provider is associated with a particular type of hospital, there is a potential for confounding of provider type with hospital type. It is important to know the extent to which hospital characteristics influence the decision on mode of delivery, compared with other health care system characteristics, so that future interventions can be effectively targeted.

Insurance

Stafford¹⁷⁰ reported on a cohort of women who delivered in 1986 in California in a study rated good. Unadjusted rates of VBAC were 8.1 percent (95 percent CI, 7.6 percent to 8.6 percent) for private insurers, 8.3 percent (7.3 percent to 9.4 percent) for non-Kaiser HMOs, 9.4 percent (8.6 percent to 10.1 percent) for Medi-Cal (California Medicaid), 18.1 percent (16.3 percent to 19.9 percent) for self-pay, 19.9 percent (18.3 percent to 21.5 percent) for Kaiser Permanente, 24.8 percent (20.4 percent to 29.3 percent) for indigent services, and 17.1 percent (10.5 percent to 19.7 percent) for other payers. Stafford reported that the unadjusted rates were similar to rates stratified on three potential confounders and rates adjusted by logistic regression model but only reported unadjusted rates. Stafford¹¹⁶ reported adjusted ORs for the above cohort in another study rated good. The adjusted ORs for VBAC compared with ERCD (with private insurance as the reference) were 1.0 (95 percent CI, 0.8 to 1.1) for non-Kaiser HMO, 0.8 (0.8 to 0.9) for Medi-Cal, 1.7 (1.5 to 1.9) for self-pay, 3.9 (3.3 to 4.6) for Kaiser-Permanente with Kaiser payment, 2.6 (1.4 to 4.6) for Kaiser Permanente without Kaiser payment, 1.9 (1.0 to 3.6) for indigent services, and 1.3 (1.1 to 1.5) for other payers. All were significantly different from the reference except for nonKaiser HMO.

King and Lahiri¹¹⁵ compared VBAC rates and adjusted ORs (adjusted for baseline risk and potential confounders) for VBAC across four insurance types. There was little variation among VBAC rates (21 percent for Medicaid to 25 percent for HMOs) and only the OR between HMOs and private insurance was different from 1 (1.15, 95% CI: 1.02, 1.30). The authors provided results for the state of New York that both included and excluded data from New York City. If data from New York City were omitted, the previous OR would not be different from 1 (OR 1.03; 95% CI, 0.90 to 1.17) but the OR comparing self-pay with private insurance (1.28; 95% CI, 1.01 to 1.81) would differ significantly from 1 (this OR was not different with data from New York City included). These ORs are all close to 1.0 whether or not they are statistically significant, suggesting a weak relationship of insurance type with VBAC rate.

A multivariable regression model by Santerre¹³⁶ showed no effect of payment source (private payer or public payer) on rates of VBAC. Thus, insurance type had no impact on VBAC rates after adjusting for other factors. Similarly, Gregory et al.¹⁶⁴ found no difference in VBAC rates for a dichotomous payment source variable (private insurance: yes or no) in a multivariable regression model.

The association between types of insurance (or payer) and VBAC rates are inconsistent across studies. While data from 1986 in California showed substantially higher rates of VBAC with Kaiser Permanente coverage and, to a lesser extent, indigent services and self-pay, similar associations have not been seen in other studies. Thus, this result may have been unique with respect to state, year, and payor.

In summary, because many factors including patient characteristics, access to obstetric providers, practice variation among providers, training of providers, ability to perform a cesarean expeditiously, and hospital characteristics may all influence the likelihood of a patient to choose TOL and the safety of each choice, current studies have not been able to identify the conditions that increase risk of TOL or ERCD. While the various characteristics of health care systems have been discussed separately above, studies need to look across these characteristics to provide a complete picture and avoid potential confounding variables. For example, an analysis of type of provider might determine a lower rate of VBAC among midwives than among obstetricians. However, midwives might be more likely to provide obstetric care to women without insurance and women of lower education levels and socio-economic status, and might be more likely to work in clinical settings without around-the-clock availability of surgical and anesthetic services and might be subject to different legal restrictions. Given the large number of potential confounders, careful adjustment for these potential confounders needs to be performed. This will require large and detailed data sets with information on patients (both mother and newborn), hospital, and provider.

Summary

• Studies of legislation, policy, guidelines, hospital characteristics, provider characteristics, insurance type or access to care focus exclusively on VBAC rates rather than safety.

• There are no studies regarding the impact of the current malpractice crisis on availability of obstetric providers and impact on a patient's options.

• Studies of provider characteristics failed to control for important confounders such as patient selection bias.

• Studies of hospital characteristics consistently report higher VBAC rates for teaching hospitals, but they conflict on whether having a NICU affects rates.

• The association between insurance status and VBAC rates is inconsistent among studies

• Current studies have not controlled for confounding for factors such as patient selection bias, as such, they have not identified conditions or practice management styles that increase risk of TOL or ERCD.