Trial of Labor Rate

In order to understand the proportion of women who have a VBAC, the first searches were for studies to estimate the proportion of women attempting a TOL in the United States (U.S.). To be included studies had to be at least fair quality, clearly define eligibility for TOL, as well as provide the number of women eligible for TOL and the number of women who had a TOL. Thirty-five observational studies consisting of 10 prospective cohort studies^76–85 and 25 retrospective cohort studies,^{10, 27, 86–108} were combined providing a TOL rate of 61 percent (95 percent CI: 57 to 65 percent). This analysis (and all meta-analyses) was conducted using a random effects model that considers heterogeneity. In this analysis of 35 observational studies that included 661,765 TOL-eligible women, the Q-statistic for heterogeneity was high, with an I² for between-heterogeneity of greater than 99 percent. The full range for TOL rates across the studies inside and outside the U.S. was 28 to 82 percent, Figure 6. Further stratification of study subgroups were conducted to assess differences that could explain the heterogeneity. Considered factors included study design (prospective versus retrospective), U.S. versus non-U.S. population, gestational age of the population (term versus any gestational age), and year of study. Among these, gestational age, country of origin, and year of study demonstrated statistically significant differences. Stratification by the study designs (retrospective and prospective) of all 35 studies revealed no significant association with TOL rates.

Figure 6

Trial of labor in studies conducted in the United States and outside the United States.

As shown in Figure 6, the overall TOL rate in studies conducted in the U.S. was 58 percent (95 percent CI: 52 to 65 percent), with a range of 28 to 70 percent, compared with 64 percent (95 percent CI: 59 to 70 percent) among women in studies conducted outside the U.S.

Similarly, fewer women in studies conducted exclusively in term populations both inside and outside the U.S. had a TOL, 53 percent (95 percent CI: 47 to 59 percent) compared with 66 percent (95 percent CI: 61 to 70 percent) for studies that included any gestational age (GA, p=0.002). Further stratification by country and by GA (shown in Figure 6) revealed that this relationship was present in both U.S. and non-U.S. studies but was only statistically significant in studies conducted outside the U.S. (p=0.001).

Because VBAC rates have changed over time¹⁰⁹ with dramatic decreases after 1996 when new evidence about uterine rupture was published, ¹⁰ the studies were grouped by the years when the data were collected: completed before 1996; included 1996; started after 1996. Before 1996, the summary estimate of the TOL rate in the U.S. was 62 percent (95 percent CI: 58 to 67 percent).^{77, 79, 81, 84, 90, 92, 94, 102, 105, 107} The summary estimate for the TOL rate in studies that were started before 1996 but completed during or after 1996 was 63 percent (95 percent CI: 55 to 72 percent).^{97, 98, 110} Finally, the summary estimate for the TOL rate for studies started and completed after 1996 was 47 percent (95 percent CI: 37 to 58 percent).^{78, 80, 87, 88, 93} Thus TOL rates among studies conducted after 1996 were significantly lower than studies conducted prior to 1996 (p=0.009) or that included 1996, p=0.036). This pattern was also observed among studies conducted outside the U.S., see Figure 7.

Figure 7

Global trial of labor rates have dropped over time. *US studies started after 1996 reported lower rates of TOL than studies that included 1996 (p=0.036) or were completed by 1996 (p=0.009) ^†Non-US studies started after 1996 reported lower rates (more...)

Almost all of the TOL studies were conducted in tertiary care centers, teaching hospitals with residents and 24-hour anesthesia teams available; therefore, findings have limited applicability to rural settings. Because stratification by setting was not feasible, studies that reported rates across settings were examined to find information for rural settings. Three retrospective studies^{10, 86, 94} reported reduced attempts (TOL rates ranging from 36 to 47 percent) for rural settings compared with urban and/or teaching settings (TOL rates ranged from 53 to 64 percent). Consistent with this finding, one retrospective study reported that 70 percent of repeat cesarean deliveries (RCD) in rural settings were potentially unnecessary compared with 61 percent in urban teaching settings.¹¹¹

Summary of trial of labor rates. The rates of TOL are highly variable ranging from 28 to 70 percent in the U.S. The evidence is largely limited to large tertiary teaching hospitals. TOL rates have declined, particularly after 1996, both inside and outside of the U.S. In the U.S. studies launched after 1996, less than half (47 percent) of women in the studies had a TOL.

Predictors of Trial of Labor

Eight good or fair quality retrospective cohort studies,^{10, 27, 86, 94, 101, 112–114} and one fair quality retrospective cross sectional study¹¹¹ looked for factors known in the prenatal setting that may predict TOL. Three themes emerged from these studies related to site of delivery, history of prior vaginal delivery, and race. TOL was more likely in hospitals with higher delivery volumes, tertiary care centers, and teaching hospitals.^{10, 86, 94, 111} Women with a prior vaginal delivery had more than double the likelihood of a TOL (odds ratio 1.51 to 6.67).^{10, 86, 101} Finally, non-white women were more likely to have a TOL than white women (odds ratio 3.5).²⁷ Further details and discussion of predictors of TOL can be found in Appendix K.

Vaginal Birth After Cesarean Rate

As the TOL rate is decreasing, it is certainly important to examine what affect, if any, this has on the VBAC rate and what factors are contributing to vaginal delivery. Sixty-seven studies, 14 fair quality prospective cohort studies^{76–85, 115–118} and 53 retrospective cohort studies^{60, 86–93, 110, 119–128 10, 27, 95–98, 100–108, 129–146} provided an overall summary estimate for VBAC of 74 percent (95 percent CI: 72 to 75 percent). The heterogeneity of this meta-analysis of 67 observational studies that included 368,304 women was high, I² greater than 98 percent. The range in VBAC rates across studies inside and outside the US was 49 to 87 percent, Appendix L. To examine this heterogeneity, these studies were stratified and analyzed by study design (prospective versus retrospective; true cohort that included TOL and ERCD versus studies of TOL only), country (U.S. versus non-U.S.), GA (term only versus any GAs) and by years when the data were collected (completed before 1996, during 1996, and started after 1996). None of these factors were found to result in statistically significant differences (see Appendix L for detailed evaluation).

Summary and strength of evidence on vaginal birth after cesarean rates. The overall strength of the body of evidence is moderate. The rates of VBAC are highly variable in these studies. Most evidence of VBAC rates are from studies based in large tertiary care centers. While TOL rates have dropped over time, VBAC rates reported in observational studies have remained constant for the women who have a TOL. In studies based in the U.S., 74 percent (95 percent CI: 72 to 76 percent) of women who had a TOL delivered vaginally.

Induction of Labor

A major area of interest is whether clinical antepartum and intrapartum management strategies such as induction of labor (IOL) influence VBAC rates. Women with a history of a prior cesarean delivery who undergo a TOL may require cervical ripening, induction and/or augmentation of labor. Multiple approaches can be taken including mechanical, pharmacological or combinations. The potential impact of each method on the ultimate VBAC rate, as well as any harm or benefit to the mother or infant is important to understand. It is particularly important when reviewing medications to consider the impact of type of drug, dose and regimen as well as important characteristics of the mother, the labor, and the fetus. Cervical ripening, which is used when the cervix is determined to be “unfavorable” for induction (generally a Bishop’s score of less than 6), can be accomplished using various mechanical methods including artificial rupture of membranes and the introduction of a Foley catheter into the cervix. The pharmacological approach to cervical ripening is generally accomplished using a prostaglandin—misoprostol (prostaglandin E₁ analog) or prostaglandin E₂—although the progestin blocker, mifepristone, has also been studied. Prostaglandins and oxytocin are used for labor induction, and oxytocin is frequently used to augment labor that is not progressing as expected.

While comparisons are sometimes made between women undergoing IOL to those with spontaneous labor, the reasons that induction is necessary may confound any effect of the drug or mechanical method of induction. However, the realistic comparison group for women undergoing IOL is expectant management as women who present in spontaneous labor are no longer candidates for IOL. Women who are induced may have different risk factors for VBAC and other outcomes; therefore, the proportion of women with VBAC associated with each method of induction is presented as the primary outcome, with comparison to spontaneous labor presented as a secondary outcome. Other maternal outcomes were reported inconsistently such that meaningful analysis of an association with use of oxytocin was not possible.

Of 328 studies screened for inclusion relating to intrapartum and antepartum factors, 27 studies were included in the IOL analysis. The others were not included due to not reporting data on induction or augmentation, reporting data on the proportion induced or augmented in the overall study population but not stratifying the results on these factors, or because they were found to be poor quality.

Any induction method. Twenty-seven fair quality studies involving 11,938 women report on the rate of VBAC among women receiving any type of IOL.^{60, 96, 118, 125, 130, 131, 139, 141, 145, 147–164} Combining these data results in a pooled estimate of VBAC with any method of IOL of 63 percent (95 percent CI: 59 to 67 percent, Table 2, see Appendix M for additional figures). Unlike other datasets in this review, these observational studies found very similar proportions of VBAC whether or not the population was limited to women with term pregnancies. Table 2 also shows the pooled estimates for the individual methods of induction, whose estimates range from a low of 54 percent with mechanical methods to a high of 69 percent with mifepristone.

Table 2

Vaginal birth after cesarean rates by types of induction.

Prostaglandins. Prostaglandins are used to ripen the cervix and to initiate labor. The two prostaglandins that can be used for this purpose are prostaglandin E₂, which is available in two formulations in the U.S. (Prepidil® gel and Cervidil® insert, both as dinoprostone), and misoprostol, which is an analog of prostaglandin E_1. In general, the prostaglandins are administered cervically or vaginally, with re-application after several hours if cervical ripening has not progressed adequately. Oxytocin may be used following cervical ripening, to augment labor, if the strength or frequency of contractions is not considered adequate to maintain sufficient labor.

Prostaglandin E₂. Nineteen studies reported the rates of women with a VBAC after receiving prostaglandin E₂ for cervical ripening for a TOL. These were a single trial in term pregnancies, one case control study, seven cohort studies including women delivering at less than 37 weeks GA, four case series of term gestations, and six case series of open GA.^{139, 145, 147–149, 151, 152, 155, 157–159, 161, 163, 165–170} Two were poor quality^{169, 170}-- inadequate reporting of numbers of women included in analyses and lack of control for potential confounding factors--while the rest were fair quality. Also, a nested case-control study did not report data adequately to determine the proportion who received PGE₂ and had a VBAC.¹⁶⁶ Some of these studies compared the women with induction with women experiencing spontaneous labor^{63, 139, 145, 149, 157, 158, 161, 163, 165–169} and others made only comparisons among induction groups.^{147, 148, 151, 152, 155, 159, 170}

Pooling studies with similar design indicates that 63 percent (95 percent CI: 58 to 69 percent) of women undergoing a TOL with IOL with PGE₂ had a VBAC. One outlier is a very small study (N=17)¹⁶³ that reported a low rate of VBAC for spontaneous labor (33 percent), but removing this study does not materially alter the results (Appendix M).

Examining the pooled analysis (Appendix M), it can be seen that including women induced prior to 37 weeks gestation (any GA) appears to reduce the rate of VBAC from 77 percent in term patients to 61 percent and 65 percent in any GA (case series and cohort studies); however the data on term gestations is limited to two case series studies. What role, if any, study design played in vaginal delivery rates was also considered. For this reasons studies are grouped in the plot according to study design. As demonstrated by the figure, vaginal delivery rates did not vary by study design.

Additional data on induction with PGE₂ come from a trial that compared weekly administration of PGE₂ (up to three doses) to expectant management. The VBAC rate with this weekly regimen was 57 percent compared to 55 percent with expectant management. Because this does not reflect usual practice with PGE₂, it may not reflect the VBAC rate found with the more traditional administration.¹⁶¹

Dose of prostaglandin was sporadically reported, and when reported was not consistent in the reporting metric (e.g., mean number of doses versus the number of women receiving two or three doses). Similarly, the number of prior cesarean delivery and the rate of prior vaginal births may have influenced the rate of VBAC with PGE₂ induction, but these covariates were not reported adequately for analysis. A small number of studies did report these variables, but either did not stratify the results by specific induction method¹⁵⁶ or had too few numbers to allow analysis.¹⁵⁷

Augmentation of labor with oxytocin may be an important confounder for VBAC and might be expected to result in improved VBAC rates. The use of oxytocin for augmentation of labor after induction with PGE₂ ranged widely in the 11 studies reporting these data, from 16 to 77 percent (Table 3). Examining the rigorousness of the study design, the inclusion of only term or any GA labors, or the size of the study does not provide further understanding of any potential relationship between these variables.

Table 3

Vaginal birth after cesarean and proportion with augmented labor in women induced with prostaglandin E₂.

Misoprostol. Evidence on the VBAC rate with cervical ripening and induction using misoprostol in women with history of prior cesarean delivery is extremely limited. The contributing evidence comes from two small (N=96 combined) fair quality retrospective cohort studies that included women with any GAs. These two studies report widely different proportions of women with VBAC, 78 percent¹⁷¹ and 44 percent,¹⁷² with a pooled estimate of 61 percent (95 percent CI: 27 to 90 percent). Although these are fair quality studies in general, they may not represent similar populations, or because they are small may not have included enough women to be able to prevent sampling error.

Mifepristone. Mifepristone is a progestin blocker that has been studied for use in cervical ripening. A small (N=32), fair quality trial compared mifepristone 200 mg and placebo, each given orally for 2 days to women with a Bishop score of three or less, followed 2 days later by induction with prostaglandins, oxytocin, and/or artificial rupture of membranes as needed. Eligible women were scheduled for IOL due to post-date pregnancy, pre-eclampsia, or severe fetal growth retardation. Although the rate of onset of spontaneous labor was higher in the mifepristone group than in the placebo group (69 versus 12 percent), the rates of vaginal delivery were not statistically significantly different (69 versus 50 percent) although the study may have not had adequate statistical power to determine a difference at a p<0.05 level.¹⁷³ Other findings that were different between the groups were the time between first dose and start of labor (60 hours and 30 minutes versus 82 hours and 50 minutes; p<0.01), and the amount of oxytocin required (2.11 international unit [IU] versus 4.67 IU; p<0.01) for mifepristone and placebo, respectively.

Oxytocin. Oxytocin is used for both IOL as well as augmentation of labor when contractions are inadequate to maintain progression of labor. Seventeen studies reported on the use of oxytocin in women with a prior cesarean delivery (one good quality cohort, nine fair quality cohorts, and five fair quality case series).^{11, 118, 125, 129–131, 141, 154, 156–158, 162, 164, 168, 169, 174, 175} While oxytocin can be used in combination with a prostaglandin, the focus here is on the use of oxytocin alone. Five studies reported VBAC rates with oxytocin alone used only for IOL (not augmentation),^{125, 141, 157, 162, 175} The pooled estimate of VBAC is 62 percent (95 percent CI: 53 to 70 percent). From the forest plot (Figure 8) it can be seen that the studies with lower quality design and including any GA report higher VBAC rates than either better studies (cohort design) or the single case series in term gestations. Additionally, the best quality study in this group reported an odds ratio for VBAC of 1.19 (95 percent CI: 0.93 to 1.53) for women who had IOL with a favorable cervix compared with women with spontaneous labor.¹⁵⁶ It is not clear that this includes only oxytocin induction, however. A fair quality study reported a VBAC rate of 74 percent, but includes oxytocin and/or amniotomy.¹⁶⁹

Figure 8

Vaginal birth after cesarean rates with oxytocin induction. Abbreviation: VBAC= vaginal birth after cesarean

Seven studies reported the rate of VBAC with oxytocin used only for augmentation.^{118, 125, 129, 131, 141, 154, 162} A fair quality case series reported the odds of a VBAC with oxytocin augmentation to be 0.83 (95 percent CI: 0.67 to 1.02) compared with no augmentation.¹²⁹ The other six studies reported data that could be pooled, resulting in a VBAC rate of 68 percent (95 percent CI: 64 to 72 percent, Appendix M).

Among studies reporting data on VBAC rate among women receiving oxytocin for induction separately from those who received oxytocin for augmentation of labor, three of four found higher proportions with VBAC when oxytocin is used for augmentation rather than induction (Table 4). Pooling these data indicates a 12 percent increase in absolute risk of VBAC when augmentation rather than induction is the reason for oxytocin use.

Table 4

Vaginal birth after cesarean rate with oxytocin used for induction versus augmentation.

Dose of oxytocin (cumulative, maximum rate of infusion) was sporadically reported, and when reported was not consistent in the reporting metric. Similarly, while the number of prior cesarean deliveries may have influenced the rate of VBAC with PGE₂ induction, these covariates were not reported adequately for analysis.

Prostaglandin E2 versus oxytocin. While the conditions that may lead to using PGE₂ for IOL may differ from those for choosing oxytocin, there may be some value in comparing the benefits and harms of these two drugs, particularly if the reason for induction and underlying obstetric characteristics also can be compared. The best evidence comparing the effect of these two drugs on VBAC rates comes from a good quality cohort study of women at term who had one prior cesarean delivery.¹⁵⁶ In this study of 11,778 women, 3,259 underwent IOL. An analysis of women induced with an unfavorable cervix compared with those experiencing spontaneous labor showed a significantly lower rate of VBAC in the induced group; odds ratio 0.46 (95 percent CI: 0.39 to 0.53). A similar analysis of women induced with a favorable cervix found no significant difference between groups; odds ratio 1.19 (95 percent CI: 0.93 to 1.53). Assuming that PGE₂ (± oxytocin) was used in those with unfavorable cervixes, and oxytocin used without PGE₂ in those with favorable cervixes, this indicates a higher rate of VBAC with oxytocin alone. However, a small proportion of women were induced without either drug and direct comparisons were not undertaken. A poor quality trial enrolled 42 women and found a higher rate of VBAC with PGE₂ (81 percent) compared with oxytocin (71 percent); however the study was too small to find a statistically significant difference and methodological concerns indicate the need to interpret these findings with caution.¹⁷⁶ Two other studies (from three publications) did not report data in a way that would allow comparison of VBAC between PGE₂ and oxytocin groups.^{164, 177, 178}

Mechanical methods of induction. The evidence on the risk or benefits of mechanical methods of cervical ripening in women with a prior cesarean delivery is limited to five studies.^{148, 150, 169, 179, 180} The best of these are two small retrospective cohort studies evaluating the use of a Foley catheter for cervical ripening compared with spontaneous labor.^{148, 150} The proportion of women with VBAC after Foley catheter IOL were 56 percent¹⁵⁰ and 51 percent¹⁴⁸ with a pooled proportion of 54 percent (95 percent CI: 49 to 59 percent).

The first is a good quality study (N=2,479), that also included a group of women with favorable cervixes who received amniotomy. Logistic regression controlling for several factors found that cervical ripening with a transcervical Foley catheter resulted in a slightly lower proportion with VBAC than with spontaneous labor (odds ratio 0.68; 95 percent CI: 0.41 to 1.15), while amniotomy resulted in a slightly higher rate (odds ratio 1.19; 95 percent CI: 0.84 to 1.69); neither was statistically significantly different from spontaneous labor.¹⁵⁰

The other study (N=1,648) included women who received PGE₂ tablets or Foley catheter for cervical ripening compared with spontaneous labor.¹⁴⁸ This study has more flaws than the first, in that the baseline characteristics, such as number or prior cesarean deliveries, previous vaginal deliveries, inter-delivery interval, etc., are not as well described; neither was the dosing of PGE₂ taken into account nor did the analysis control for confounding variables. This study found the rate of VBAC to be significantly lower in the Foley catheter group compared with the spontaneous labor group (51 versus 65 percent, p< 0.01).

In the first study, crude rates of VBAC were also lower in the Foley catheter group, but analysis controlling for confounding factors resulted in a non-significant difference.¹⁵⁰ The authors of both studies note that for the outcome of uterine rupture, the studies were most likely underpowered to find differences between the groups; they estimate that it would require a sample size of greater than 10,000 patients to find a difference. Additional evidence on Foley catheter, a double-balloon device, and breast stimulation were inadequate to make determinations of VBAC rate, risk of uterine rupture, or other outcomes.^{169, 179, 180} No studies of other mechanical methods, e.g., acupuncture, were found.

Influence of prior vaginal delivery and indication for prior cesarean on vaginal birth after cesarean induction. A small case series of women requiring oxytocin induction or augmentation (N=98) did not find a significant difference in the rate of VBAC among those with prior vaginal deliveries compared with those with none among the group induced with oxytocin, but did find that prior vaginal delivery resulted in a higher rate of VBAC among those with oxytocin augmentation compared to those with no prior vaginal deliveries (86 and 56 percent, p<0.05).¹⁶² In this study, a recurrent indication for induction or augmentation resulted in significantly lower rates of VBAC compared with the groups without recurrent indications (38 versus 90 percent, p<0.01 and 81 versus 50 percent, p<0.05, respectively). Comparing the birth weights of current and previous deliveries in the induced and augmented groups resulted in a non-significant difference with those induced, but a difference favoring smaller fetuses among those augmented (84 versus 49 percent, p<0.01). Because this is a small case series, these data should be interpreted as suggestive, rather than conclusive.

In a fair quality study of 205 women with a prior cesarean delivery and who required induction, 41 percent of those with no prior vaginal deliveries had a VBAC, while 82 percent of those who did have prior vaginal delivery had a VBAC (odds ratio 6.8; 95 percent CI: 3.04 to 13.9).¹⁵⁷

The relationship between the reason for the prior cesarean delivery and VBAC when oxytocin is used was examined in three studies (total N=595).^{118, 125, 130} The most commonly reported indications and pooled proportions with VBAC are listed in Table 5 below.

Table 5

Most commonly reported indications and pooled proportions.

Summary and strength of the evidence on induction of labor. Overall, evidence regarding the rate of VBAC among women with IOL is low to moderate strength, indicating that 63 percent of these women will have a VBAC (PGE₂=63 percent, oxytocin=62 percent, misoprostol=61 percent). Augmentation of labor with oxytocin was associated with a rate of 68 percent VBAC, although the strength of this evidence is low. Evidence was inadequate to make a comparison of VBAC resulting from induction with oxytocin or prostaglandin E₂. Fifty-four percent of women induced with a Foley Catheter had a VBAC, based on moderate strength evidence. Other mechanical methods were not reported adequately to make comparisons or conclusions. Approximately 60 percent of the studies were conducted in university hospital settings, with the smaller studies also including community hospital settings such that the evidence is weighted towards the larger, tertiary care setting. Less than half of these studies were conducted in the U.S., with most of the others being conducted in Canada, Britain, Ireland, Sweden and Israel. For induction, the results were not stratified by age, race, ethnicity, or baseline risk.

Predictors of Vaginal Birth After Cesarean

The impact of individual factors on VBAC discussed above can overlap and interact with each other such that a factor found to have statistically significant influence on VBAC rate may no longer be significant when other key factors are taken into account. Studies that evaluate these factors in concert, using regression analyses for example, can provide a higher level of evidence on the residual influence of individual factors. Four prospective cohort studies,^{85, 167, 181, 182} 18 retrospective cohort studies,^{27, 61, 86, 99, 104, 106, 113, 119, 127, 136, 143, 146, 183–188} and one case-control study¹⁸⁹ addressed predictive factors for VBAC. The key factors considered by these studies were demographic factors that included maternal age, ethnicity, race, and marital status; nonclinical factors that included insurance status, site of delivery, and volume of VBACs; past obstetric factors that included prior vaginal delivery and prior indications for cesarean delivery; pre-existing and current factors that included maternal height, body mass index (BMI), substance abuse, and pre-existing maternal disease; and current obstetric factors related to the infant gender, age and size. In this section, only the strongest predictors (supported by several good or fair cohort studies) that had an overall trend are presented. Full discussion of all predictors can be found in Appendix K.

Demographic factors. Of all demographic predictors that were evaluated, the strongest evidence was found for ethnicity and race. In all four cohort studies reporting on ethnicity and race, Hispanic women had a reduced likelihood of VBAC (by 31 to 49 percent, Appendix K) than non-Hispanic women.^{181, 182, 185, 187} In these same studies, African American women had a reduced likelihood of VBAC (by 20 to 49 percent) compared with white women.^{181, 182, 185, 187} It is interesting to note that non-white women were more likely to have a TOL but less likely to have a VBAC.²⁷ While not included in the meta analysis of VBAC rates because the cohort included twin deliveries, one retrospective cohort study reported that African American women were more likely to fail at VBAC, were more likely to be hypertensive and have diabetes.¹⁹⁰

Nonclinical factors. The site of delivery can play an important role in deciding what kind of birthing options are pursued, (Appendix K). Women at rural and private hospitals that provide obstetric care for lower risk deliveries had a decreased likelihood of VBAC.^{86, 111} This finding is consistent with another finding by the same investigators⁸⁶ that women at rural and private hospitals were less likely to attempt a TOL. Private hospitals in this study had an average VBAC rate of 57 percent compared with a VBAC rate of 66 percent for perinatal centers.⁸⁶

Past obstetric factors. There is particular interest in whether demographic factors, nonclinical and past obstetric factors may predict VBAC since these factors are known prenatally and would allow clinicians to provide information on prognosis early in pregnancy. Investigators from studies have explored prior vaginal delivery, years since prior cesarean delivery, prior labor experience and prior baby weight as potential factors for predicting VBAC. A prior history of vaginal delivery was consistently reported to increase likelihood of VBAC in all 13 cohort studies^{86, 99, 104, 127, 136, 143, 167, 181–184, 187, 191} and one case-control study (Appendix K).¹⁸⁹ Women with a vaginal delivery after their prior cesarean (prior VBAC) were three to seven times more likely to have a VBAC for their current delivery^{127, 167, 181–184, 189, 191} compared with women with no prior vaginal deliveries. Women who had a vaginal delivery before their cesarean deliveries also had an increased likelihood to have a VBAC.^{143, 167}

One secondary analysis of a retrospective cohort study⁹⁸ of 16 community and university hospitals¹²³ specifically examined the effect of prior vaginal delivery before a cesarean and of a prior VBAC on the current TOL. The VBAC rate for women with no history of vaginal delivery was 65 percent, 83 percent for women with a prior vaginal delivery before a cesarean, and 94 percent for women with a prior VBAC.¹²³ Finally, a secondary analysis of the National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network (MFMU) cohort data reported that the likelihood of a VBAC increased with each prior VBAC.¹⁹² Women with zero, one, two, three, and four or more prior VBACs had likelihoods of VBAC of 63.3, 87.6, 90.9, 90.6, and 91.6 percent (p<0.001), respectively.

Pre-existing and current obstetric factors. Many pre-existing factors (maternal height, BMI, smoking and substance use, and maternal disease) appear to change the likelihood of VBAC (Appendix K) but were only supported by single or two studies and are discussed in the appendix. Several investigators examined the effect of pre-existing disease on VBAC. In three of four cohort studies, women with a maternal disease (hypertension, diabetes, asthma, seizures, renal disease, thyroid disease, or collagen vascular disease) had a decreased likelihood (by 17 to 58 percent) of VBAC.^{127, 182, 187} By contrast, in a large prospective study by Grobman et al,¹⁸¹ the presence of diabetes, asthma, chronic hypertension, renal disease or heart disease, was not significant in the study’s multivariable logistic model.

Many current obstetric factors related to the infant (GA, birth weight and infant gender) predicted VBAC or TOL followed by a cesarean delivery. Of these, the most consistent finding is that as infant weight increases the likelihood of VBAC decreases. Four of five studies reported that women delivering infants weighing more than 4 kilograms (kg) had a reduced likelihood of VBAC (by 41 to 51 percent) than women who delivered smaller infants.^{86, 127, 143, 146, 182} The oldest of these studies¹⁴³ found no relationship between a birth weight over 4 kg and likelihood of VBAC.

Obstetric factors related to the labor itself (dilation, effacement, station, Bishop score, cervix position) consistently predicted VBAC. Three prospective cohort studies,^{167, 182, 184} one retrospective study¹³⁶ and one case-control study¹⁸⁹ provided consistent evidence that women who were more dilated at admission or at rupture of membranes (ROM) were more likely to deliver vaginally. All three studies that examined effacement reported increased likelihood as effacement reached 75 to 100 percent.^{99, 167, 184} Similarly, all three studies that examined head position reported that as the baby’s position was vertex, engaged or at a lower station, the likelihood of VBAC increased.^{85, 136, 185} Both studies that examined Bishop’s score showed that as the score increased, the likelihood of VBAC increased two¹⁸³ to six times.¹⁴³

Overall impression of prediction studies. With the exception of three studies,^{104, 167, 181} these prognostic studies of VBAC could be described as exploratory.¹⁹³ According to Simon and Altman, studies that report association and identify patients at risk but that have not yet had results confirmed in followup studies with pre-stated hypotheses, do not yet provide sufficient evidence to change clinical practice.¹⁹³ The three studies^{104, 167, 181} that provided this cross‐validation evidence also proposed screening tools for VBAC and are discussed in detail in Appendix N.

Summary of predictors of vaginal birth after cesarean rate. Hispanic and African American women were more likely to have a TOL but less likely to have a VBAC compared with non-Hispanic and white women, respectively. Women at rural and private hospitals had a decreased likelihood of TOL and a decreased likelihood of VBAC. A prior history of vaginal birth was consistently reported to increase likelihood of VBAC. Women delivering infants over 4 kg have a reduced likelihood of VBAC. Greater progress of labor--measured as greater dilation, lower station and higher Bishop score--predicted a higher likelihood of VBAC.

Screening Tools for Predicting Vaginal Birth After Cesarean

The purpose of a screening tool is to help providers and patients to better identify who will have a VBAC (and who is more likely to have a RCD). Currently, most women are told they have a likelihood of 60 to 80 percent for a VBAC.¹⁹⁴ Screening tools are most helpful for women who have an estimate of VBAC outside this range either to discourage or to encourage a TOL. Two prospective cohort studies,^{167, 181} 10 retrospective cohort studies^{104, 107, 116, 120, 122, 128, 142, 143, 183, 184} and two case-control studies^{102, 189} that presented screening tools were identified (Appendix N). These studies combined individual factors to predict the likelihood of VBAC (or RCD) when certain thresholds were reached. Predictive variables (historic, intrapartum or perinatal) of delivery route were first identified by univariate analyses. Significant variables (p<0.05) were included in multiple logistic regression models and/or scored models.

In the strongest studies, the resulting models or scoring systems were then evaluated with a separate validation dataset.¹⁹⁵ Three of the scored models had one or more external validation studies that tested the models with independent cohort datasets. The scored model by Flamm¹⁶⁷ was externally validated by one retrospective cohort study.¹²² The scored model by Grobman 2007¹⁸¹ was externally evaluated with a retrospective study.¹²⁰ The Troyer¹⁰⁷ model was externally validated by two retrospective studies.^{122, 142} In all validation studies, the scored model’s performance was similar to the originally reported performance (see shaded rows of Table N-1 in Appendix N). In a retrospective cohort study that evaluated three scored models using the same dataset,^{107, 167, 196}, Dinsmoor et al reported that all three models were accurate at predicting which women would have a VBAC but were not accurate at predicting who would have a RCD after a TOL.¹²² Using the three models, 50 percent of women with unfavorable risk factors had vaginal deliveries, suggesting that other factors may be needed to identify women at risk for cesarean delivery. A previous decision analysis of VBAC¹⁹⁷ suggested that a scored model would be most useful clinically if it achieved a sensitivity and specificity greater than 85 percent,¹²² which none of these tools achieved. All scored models are presented and discussed in detail in Appendix N.

Summary of screening tools. Since the VBAC evidence report published in 2003^{62, 194} five scored models have been created and evaluated to identify women for VBAC (or for RCD).^{104, 128, 181, 183, 184} Two of the studies created scored tools that can be used in the prenatal setting.^{128, 181} All scored models provide reasonable ability to identify women who are good candidates for VBAC but none have discriminating ability to consistently identify women who are at risk for RCD.

Maternal Death

Maternal mortality rose in the U.S. after the dawn of the 21^st century from seven to nine per 100,000 in the 80s and 90s to 12 to 15 per 100,000 since 2003.^198–200 Although these rates are still very low, understanding whether choices patients and providers make about route of delivery and clinical management contribute to maternal mortality is critically important.

There were 12 good or fair quality cohort studies totaling 26 maternal deaths among 402,883 patients that reported maternal mortality associated with TOL and ERCD (Appendix O).^{76, 77, 81, 93, 97, 100, 108, 164, 201–204} The absolute risk of maternal death associated with TOL was 0.0038 percent (95 percent CI: 0.009 to 0.0155 percent) and with ERCD was 0.0134 percent (95 percent CI: 0.0043 to 0.0416 percent). Four studies focused exclusively on women delivering at term.^{93, 97, 108, 204} Among these four studies the maternal mortality was similarly lower for TOL (0.0019 percent; 95 percent CI: 0.0004 to 0.0095 percent) compared with ERCD (0.0096 percent; 95 percent CI: 0.0021 to 0.0432 percent, Figures 9 and 10).

Figure 9

Rates of maternal death for trial of labor among all studies. *95% confidence intervals are exact Abbrevation: TOL=trialof labour

Figure 10

Rates of maternal death for elective repeat cesarean delivery among all studies. *95% confidence intervals exact Abbrevation: ERCD=elective repeat cesarean delivery

One study from Canada reported differences in maternal mortality with TOL relative to ERCD for low and high volume maternity units. This study deserves mention because it was one of the only studies to evaluate maternal death in a range of institution types. This study evaluated both large tertiary centers as well as community hospitals, demarcating types by the number of deliveries per year.¹⁰⁸ In low volume maternity wards defined as less than 500 deliveries a year, the odds ratio was 2.68 (0.16 to 45.5) for maternal mortality with TOL compared with RCD. High volume maternity wards, defined by greater than 500 deliveries per year, were noted to have an odds ratio of 0.16 (0.02 to 1.29) of mortality with TOL compared with RCD.¹⁰⁸ No other studies stratified maternal death rates by institution size or delivery volume.

Summary and strength of evidence on maternal death. Overall, the strength of evidence regarding the rate of mortality for women with a prior cesarean delivery is high with good consistency and precision. While maternal mortality is rare with an overall rate of 10.1 per 100,000 for all women with prior cesarean, the risk of maternal mortality is significantly increased with ERCD. When combining the TOL group across all studies, the risk of maternal mortality is found to be 0.0038 percent (95 percent CI: 0.0009 to 0.0155 percent). The combined risk for ERCD group across all studies is 0.0134 percent (95 percent CI: 0.0043 to 0.0416 percent). This translates to 3.8 per 100,000 for TOL (95 percent CI: 0.9 to 15.5 per 100,000) and 13.4 per 100,000 for ERCD (95 percent CI: 4.3 to 41.6 per 100,000). When the analysis is limited to term studies, the combined risk of maternal deaths for TOL is 0.0019 percent for TOL (95 percent CI: 0.0004 to 0.0095 percent) and 0.0096 percent for ERCD (95 percent CI: 0.0021 to 0.0432 percent); translating to 1.9 per 100,000 for TOL (95 percent CI: 0.4 to 9.5 per 100,000) and 9.6 per 100,000 for ERCD (95 percent CI: 2.1 to 43.2 per 100,000). In addition, meta‐regression showed that among TOL studies, maternal mortality was significantly lower among studies of term populations compared with studies of any gestational age (p=0.027) but there was no significant difference based upon gestational age among ERCD studies (p=0.141) (Figures 9 and 10).

While rare for both TOL and ERCD, compared to ERCD, the overall risk of maternal death associated with TOL is significantly lower (RR, 0.33, 95 percent CI: 0.13 to 0.88; p=0.027). Using 0.0134 percent as the baseline risk for ERCD, the calculated risk difference is −0.0090 percent (95 percent CI: −0.0117 to 0.0016 percent), translating to 9.0 less deaths per 100,000 (95 percent CI: 1.6 to 11.7 less deaths per 10,000) from the TOL group. Among the four studies focused exclusively on women delivering at term, the maternal mortality risk was similarly lower for TOL (RR: 0.27, 95 percent CI: 0.09 to 0.85; p=0.025). Using 0.0096 percent as the baseline risk for ERCD, the calculated risk difference is −0.0070 percent (95 percent CI: −0.0087 to 0.0014 percent), translating to 7.0 less deaths per 100,000 (95 percent CI: 1.4 to 8.7 less deaths per 10,000) from the TOL group, as compared to the ERCD group.

Uterine Rupture

Uterine rupture is a potentially life-threatening complication that has been directly attributed to VBAC. By itself, uterine rupture – defined as complete separation through the entire thickness of the uterine wall (including serosa) – is a visible or palpable anatomic finding rather than a health outcome. However, its association with perinatal and maternal morbidity and mortality raises substantial concerns among patients, clinicians, hospitals, and policymakers. Given this, there is considerable interest in what populations and conditions make VBAC a reasonable option and what if any management factors may reduce either the occurrence of uterine rupture or the severity of consequences. This section summarizes studies reporting on the risk of uterine rupture for women with TOL and ERCD, the risk of perinatal morbidity and mortality associated with uterine rupture, management factors that may contribute to the development of uterine rupture or severity of morbidity or mortality associated with uterine rupture, and techniques and tools proposed to stratify populations of women with prior cesarean for risk of uterine rupture.

Risk of uterine rupture. Estimating the risk of uterine rupture for women with a prior cesarean has been challenging not only because studies report on actual rather than intended route of delivery but also because studies often mixed true anatomic ruptures with asymptomatic dehiscences. While numerous studies have been published relating to uterine rupture and/or dehiscence (393 articles), only eight cohort studies^{10, 97, 119, 204–208} were good or fair quality, included the population of interest, and used the anatomic definition for uterine rupture contained in this report (See Appendix F for a table of excluded studies). Details of the eight included studies that included cohorts of either TOL alone or TOL and ERCD together are presented in Table 6.

Table 6

Uterine rupture: trial of labor versus elective repeat cesarean delivery among any gestational age studies.

Four studies reported uterine rupture outcomes for both TOL and ERCD.^{10, 97, 204, 205} Among these four studies totaling 47,202 patients, there were 154 uterine ruptures; 96 percent (N=148) of which were incurred by the TOL group. As shown in Figure 11, these four studies indicate that the risk of uterine rupture for all women with a prior cesarean delivery regardless of route of delivery is 0.3 percent (95 percent CI: 0.23 to 0.40 percent).

Figure 11

Uterine rupture among studies including both trial of labor and elective repeat cesarean delivery. *95% confidence intervals are exact † Test for heterogeneity is based on logits of rates

Within these four studies, the combined risk of uterine rupture for women undergoing a TOL is 0.47 percent (95 percent CI: 0.28 to 0.77 percent) and 0.026 percent (95 percent CI: 0.009 to 0.082 percent) for women undergoing an ERCD. The increased risk for uterine rupture among the TOL group is largely affected by the Spong et al study, which reports an occurrence of uterine rupture that is double that reported for the other three studies (0.7 percent versus 0.03 to 0.4 percent).^{10, 97, 204, 205} Table 7 presents details regarding the four studies that report uterine ruptures related to both TOL and ERCD to understand factors that may explain the increased risk. The Spong study is the only one to include women with incisional types other than low transverse cesarean delivery (LTCD). However, given the small contribution of these other scar types to the overall dataset, and the fact that the uterine rupture occurrence among women with LTCD in this study was also higher than the other studies (0.75 percent) it is unlikely that this alone explains the difference. Fewer women with a prior cesarean delivery elected TOL in this study at 39 percent. It is difficult to assess whether the years that studies were conducted affected the occurrence of uterine rupture because the two studies conducted after 1996 have very different populations.^{204, 205} Unfortunately, none of the four studies provided details on the proportion of women in the study who underwent IOL, a factor that is known to have large variation and to increase the occurrence of uterine rupture. The effect of IOL upon uterine rupture is considered in detail later in this section. Overall, there is no clear factor that is associated with higher versus lower occurrence of uterine rupture among the four studies providing comparative data.

Table 7

Characteristics of uterine rupture studies of trial of labor and elective repeat cesarean deliver.

Four additional studies reported on uterine rupture exclusively in women undergoing TOL.^{119, 206–208} A sensitivity analysis was performed to examine whether this difference in cohort assembly (TOL only versus TOL plus ERCD) affected the results for uterine rupture and there was not a statistically significant difference. The increased number of studies available for TOL allows a more detailed examination of factors such as gestational age, direction of scar, year of study, etc., that may be associated with higher risks of uterine rupture.

As shown in Figure 12, the occurrence of uterine rupture for TOL remains relatively unchanged at 0.46 percent. Among TOL studies, the occurrence of uterine rupture is significantly higher for studies limited to term patients compared with studies including patients of any GA (0.78 versus 0.32 percent, p=0.033). Looking across all eight studies, the highest occurrence of uterine rupture among TOL patients were reported by studies that included women with any direction of cesarean scar.^{119, 204, 206} Among these, only Spong et al presented details for uterine rupture occurrence according to the direction of incision, with the lowest rate reported in the unknown incision group 0.63 followed by LTCD 0.75, classical T or J 1.59 and low vertical 2.47. There were no uterine ruptures among women who experienced ERCD without labor for any direction of incision. This study also provides additional information relative to presence or absence of labor and indication for cesarean delivery among the groups because of its unique study design. Because it is clinically important to understand the additional risk for uterine rupture given both the number and direction of prior cesareans, studies that specifically addressed this question are presented in further detail later in this report under the section entitled “Special Considerations.”

Figure 12

Uterine rupture among all trial of labor studies. *95% confidence intervals are exact ^†All test for heterogeneity are based on logits of rates

Morbidity and mortality related to uterine rupture. As shown in Table 8, there were no maternal deaths due to uterine rupture in any of the eight studies.^{10, 97, 119, 204–208} The risk for perinatal death in the event of uterine rupture ranged from 0 to 20 percent with a pooled risk of 6.2 percent and the highest risk experienced by the TOL group. For perinatal death, it is useful to limit to term studies to remove the effect of gestational age. Among the three term studies,^{97, 119, 204} two provided data for the risk of perinatal death given uterine rupture.^{97, 119} They report that 0 to 2.8 percent of all uterine ruptures resulted in a perinatal death. Figure 13 presents uterine rupture related perinatal death among the six studies reporting uterine rupture for populations of women delivering at any gestational age.^{10, 97, 119, 206–208} Four of the eight studies reported the risk of hysterectomy given uterine rupture;^{10, 207–209} only one of which provided comparative data between TOL and ERCD groups.¹⁰ They reported an occurrence of hysterectomy give uterine rupture of 14 to 33 percent.

Table 8

Rupture associated morbidity.

Figure 13

Risk for perinatal death given uterine rupture. *95% confidence intervals are exact

Summary and strength of the evidence on risk of uterine rupture. Overall, evidence regarding the rate of uterine rupture for women with a prior cesarean delivery is moderate in strength, indicating that the risk of uterine rupture for women with prior cesarean is 0.3 percent. Compared with women undergoing an ERCD, women undergoing a TOL have a significantly higher risk of uterine rupture (RR 20.74, 95 percent CI: 9.77 to 44.02; p<0.0010). Using 0.026 percent as the baseline risk for ERCD, the calculated risk difference is 0.51 percent (95 percent CI: 0.23 to 1.12 percent) translating to 5.1 additional ruptures per 1,000 (95 percent CI: 2.3 to 11.2 per 1,000) women undergoing TOL. Though the studies are limited in number, there does not appear to be a reduction in the occurrence of uterine rupture in recent years (e.g., after 1996). Overall, studies focused on providing rates of uterine rupture but did not provide important details that would provide insights for management—such as the relationship between length of labor and uterine rupture—to establish whether there is a dose response for labor. To date, there have been no maternal deaths reported because of uterine rupture, and the risk of perinatal death due to uterine rupture is similarly low at 6.2 percent. However, the risk of hysterectomy due to uterine rupture is an important consideration for women planning VBAC, with rates ranging from 14 to 33 percent. There appears to be an increased risk for uterine rupture among women who undergo a TOL at term. Because term may also include postdates and inductions, the section that follows provides further details on the impact of management and GA upon risk for uterine rupture.

Effect of induction of labor on uterine rupture. Women often need interventions such as induction and/or augmentation of labor that may affect a woman’s risk for uterine rupture. Seven fair quality studies (four cohort, three case series) including 5,276 women with a prior cesarean delivery who had IOL report uterine rupture using the definition of separation through the entire thickness of the wall including visceral serosa (with or without extrusion of part of all of fetal-placental unit); two studies of PGE₂,145, 151 one of foley catheter used for cervical ripening,¹⁵⁰ and four of any IOL method.^{60, 96, 156, 164} However, one of the studies⁶⁰ used a definition of rupture of extrusion of the uterine contents into the peritoneal cavity, and thus likely underestimates the rate of rupture as defined in this report.

These studies indicate that the risk of uterine rupture is 1.5 percent in women receiving IOL and delivering at term and 1.0 percent when women with any GA and receiving IOL are included. These rates are two times greater compared with all women with a prior cesarean delivering at term (1.5 versus 0.7 percent), and three times greater when considering women with any GA (1.0 versus 0.3 percent). Further stratification of the data indicates that there is increased risk of rupture in women delivering at greater than 40 weeks GA (Figure 14) compared with women delivering at term (3.2 versus 1.5 percent). The reason for induction, the dose of induction agents needed, etc., need to be examined more fully to determine the cause of this increased risk. The pooled risk of uterine rupture in the spontaneous labor groups in these studies is 0.8 percent (95 percent CI 0.7 to 1.1), which is higher than the pooled estimate from all TOL studies (0.47 percent; 95 percent CI 0.28 to 0.77 percent), indicating that the baseline risk of rupture is higher in the induction studies overall. This is an important factor, suggesting that indirect comparisons from these studies to the general TOL studies is not possible.

Figure 14

Uterine rupture with induction of a trial of labor.

In women delivering at term, the risk of rupture is not statistically significantly greater in women undergoing IOL compared with those with spontaneous labor (odds ratio 1.42; 95 percent CI: 0.57 to 3.52). In examining these studies further, it appears that women at greater than 40 weeks GA have increased risk of rupture while those at term or less than 37 weeks GA do not (Figure 15). Because there were no ruptures in either group in one study⁶⁰ an odds ratio could not be calculated, but the absolute difference in risk shows that only women delivering at greater than 40 weeks GA have an increased risk with IOL over spontaneous labor (risk difference 1.8 percent; 95 percent CI: 0.1 to 3.5 percent). The number needed to harm (NNH) in this group is 56 (for every 56 women greater than 40 weeks GA whose labor is induced during a TOL, one additional rupture will occur compared with those having spontaneous labor). However, for women with indication for induction of labor at 40 weeks gestation and beyond, clinicians are faced with the option of induction or ERCD and not spontaneous labor. A better, more useful, comparator for induction of labor would be expectant management. These findings should be interpreted with caution, as one study found similarly increased rates of rupture with induction regardless of GA, while the other found increased risk in the group with greater than 40 weeks GA only.

Figure 15

Risk of rupture: induction versus spontaneous labor.

Given that there is such limited evidence regarding the risk of uterine rupture with various methods of IOL, it was decided to evaluate studies that used a broader definition of uterine rupture. While these cannot be compared with those evaluated for the risk of uterine rupture in spontaneous labor, they can be used to make indirect comparisons across IOL methods. The studies included below report that uterine rupture was a primary outcome measure of the study and report clear methods for ascertaining rupture, e.g., individual chart review. Because the definitions used are not limited to the anatomical description of rupture used in the above, these analyses should be considered exploratory only.

Prostaglandin E2. The risk of uterine rupture with PGE2 use for cervical ripening and IOL was reported as a main of outcome interest in 14 studies (six good or fair quality cohort studies, and eight fair quality case series). Two of which included only women at term gestation,^{156, 164} while the rest included women with any GA.^{11, 145, 148, 149, 151, 157, 158, 163, 165, 168–170} Pooled analysis of the proportion of women experiencing a uterine rupture after PGE2 induction provides a point estimate of 2.0 percent (95 percent CI: 1.1 to 3.5 percent, see Figure 16). It is important to note however, that these studies did not define uterine rupture according to the anatomic definition in this report. Therefore, understanding these rates in comparison with other VBAC studies is not possible. The two studies restricting to women with term GA found very different rates; 3.9 percent¹⁶⁴ compared with 0.8 percent.¹⁵⁶ The study with the lower rate was methodologically superior. The studies with more rigorous design report a higher rate of uterine rupture. One study included in this analysis used ICD-9 codes to identify uterine rupture¹¹ – a method that has subsequently been shown to overestimate true rupture rates.²¹⁰ The study reporting the highest rate, 10.3 percent,¹⁷⁰ identified cases by discharge diagnosis of uterine rupture, another method that may overestimate the proportion with true rupture. The validity and reliability of the method used was not tested or reported, and similar to the use of ICD-9 codes, could overestimate true rupture. Additionally, baseline obstetric characteristics of women induced, mean total dose of prostaglandin, and percent receiving oxytocin for augmentation (except for the uterine rupture cases) were not reported such that comparisons to other study populations cannot be made.

Figure 16

Uterine rupture with prostaglandin E₂ induction. *95% confidence intervals are exact

Dose of PGE₂ was sporadically reported, and when reported was not consistent in the reporting metric (e.g., mean number of doses versus the number of women receiving two or three doses). Similarly, while the number of prior cesarean delivery and the rate of prior vaginal deliveries may have influenced the rate of VBAC with PGE₂ induction, these covariates were not reported adequately for analysis. A small number of studies did report these variables, but either did not stratify the results by specific induction method,¹⁵⁶ or had too few numbers to allow analysis.¹⁵⁷

Misoprostol. Evidence for maternal harms with misoprostol is limited to three fair quality cohort studies, including women with any GA.^{11, 172, 211} The largest of these,¹¹ used ICD-9 codes to identify uterine ruptures, a procedure now known to overestimate rupture rates. Additionally, although the study is large (N=20,525), only 366 received prostaglandins of any kind. The data were collected between 1987 and 1996, and it is noted that misoprostol was not used regularly until 1996, indicating that few of the women described as receiving a prostaglandin had misoprostol. While the actual numbers of women receiving misoprostol are not reported, the relative risk for a uterine rupture in women who received a prostaglandin during 1996 was 12.2 (95 percent CI: 3.4 to 39.6). This compares with a separate analysis of uterine rupture among women who received a prostaglandin during the years 1987 to 1995, relative risk 14.1 (95 percent CI: 6.1 to 33.0; both compared with ERCD).

The other two studies are much smaller, but do report misoprostol data separately. In a small, fair quality, cohort study (N=226) comparing PGE₂ (administered as a gel or pessary) with misoprostol (25 to 50 mcg),²¹¹ only 16 of 145 women in the misoprostol group and nine of 81 in the PGE₂ group had a prior cesarean delivery. Additionally, the data for PGE₂ was obtained retrospectively while the data for misoprostol were obtained prospectively. The results for uterine rupture were stratified by history of a prior cesarean delivery with two of 16 in the misoprostol group (13 percent) and zero of nine in the PGE₂ group having what is described only as a low transverse rupture. One additional patient in the misoprostol group had a scar dehiscence. In both cases of rupture, hysterectomy was performed, with no blood transfusions. The second study of misoprostol made comparisons among women with a history of prior cesarean delivery to those without prior cesarean delivery, a comparison that is not relevant to this review.¹⁷² There were no uterine ruptures, and 2.0 percent were reported as having “complications,” including abruption placenta, retained placenta, uterine atony, and blood transfusion.

Mifepristone. In a small (N=32), fair quality trial of mifepristone and placebo, each given for 2 days followed 2 days later by IOL with prostaglandins, oxytocin, and/or artificial rupture of membranes as needed in term GAs,¹⁷³ one uterine scar separation occurred in each group (6.25 percent). Maternal outcomes were not different between groups, except that two patients in the mifepristone group developed fever, compared with none in the placebo group. One wound infection was found in each group.

Oxytocin. The risk of uterine rupture following induction or augmentation of labor with oxytocin was reported as a main outcome measure in eight studies (one good quality cohort study, four fair quality cohort studies, and three fair quality case series).^{11, 125, 156–158, 164, 168, 169} The risk of uterine rupture is 1.1 percent (95 percent CI: 0.9 to 1.5 percent) when data from these studies are pooled (Figure 17).

Figure 17

Uterine rupture with oxytocin. *95% confidence intervals are exact Abbreviation: GA= gestational age

A case control study of 24 cases of uterine rupture where oxytocin had been given during TOL after prior cesarean delivery and 96 controls that also received oxytocin but had no uterine rupture examined the relationship of oxytocin dose to rupture.¹⁷⁴ The study was powered to find a difference of 40 percent in the duration of oxytocin or a 65 percent increase in total dose. None of the multiple analyses found a statistically significant difference, although the difference in duration of oxytocin (530 minutes in the uterine rupture group compared with 476 in the non-rupture group) achieved a p value of 0.08. The study was small, and the large differences of 40 percent for duration and 65 percent for dose appear to have been set arbitrarily. Further analysis of the impact of dose on uterine rupture rate may be warranted.

Oxytocin augmentation of labor. Augmentation of labor with oxytocin may be an important confounder for uterine rupture; among the 12 studies reporting uterine rupture as a main outcome measure, eight reported the proportion of women receiving augmentation of labor ranging from 16 to 81 percent (Table 9).^{145, 148, 149, 151, 156–158, 168} Meta-regression of these data did not result in oxytocin augmentation to be a statistically significant covariate for uterine rupture with PGE₂ induction (p=0.22). While this evidence may be too limited to make conclusions about an association between proportions of women having both PGE₂ induction and oxytocin augmentation of labor and increasing uterine rupture rates, there is at least a trend towards increased risk when both drugs are used.

Table 9

Effect of oxytocin augmentation on rate of uterine rupture in women receiving prostaglandin E₂ for induction.

Prostaglandin E₂ versus oxytocin. The best evidence on the risk of uterine rupture with PGE₂ compared with oxytocin when used for IOL comes from a large good quality cohort study of women with term gestations and one prior cesarean delivery.¹⁵⁶ In this study there were no uterine ruptures in the prostaglandin only group, 29 out of 2,421 (1.2 percent) in the oxytocin only group, and six out of 614 in the prostaglandin plus oxytocin group (1 percent). Statistical comparisons were made only with the spontaneous labor group and stratified by prior vaginal delivery or no prior vaginal delivery. The group with no prior vaginal delivery and receiving oxytocin only resulted in a statistically significantly greater risk of uterine rupture (odds ratio 2.19; 95 percent CI: 1.28 to 3.76). Analysis of the other groups compared with spontaneous labor, including PGE₂ only, did not result in significantly increased risk. A second, lower quality cohort study similarly found the risk of uterine rupture to be significantly increased with oxytocin induction (odds ratio 4.6; 95 percent CI: 1.5 to 14.1) but not with PGE₂.¹⁶⁴ The pooled analysis of these results in an odds ratio of 2.7 (95 percent CI: 1.4 to 5.1) compared with spontaneous labor. Direct comparisons are not available, and the number of women in the prostaglandin only groups is much smaller than in the oxytocin only groups. Using the data presented in the large cohort study for uterine rupture in the oxytocin only group (prior and no prior vaginal delivery combined) compared with prostaglandin only (prior and no prior vaginal delivery combined) yielded an unadjusted odds ratio of 0.29 (95 percent CI: 0.04 to 2.09) indicating no statistically significant difference in risk. However, this is an exploratory analysis and should be interpreted with caution.

Mechanical methods of induction. The evidence on the risk or benefits of mechanical methods of cervical ripening in women with a prior cesarean delivery is very limited.^{148, 150, 169, 179, 180} The best of these are two small retrospective cohort studies evaluating the use of a foley catheter for cervical ripening compared to spontaneous labor.^{148, 150} No cases of uterine rupture occurred in the groups who had foley catheter cervical ripening, although the numbers of patients in the foley catheter groups may have been too small to identify a rupture (N=416). Additionally, while one study was rated good quality and defined uterine rupture clearly, it is not clear that evidence of uterine rupture was routinely sought in all women,¹⁵⁰ and the other study provided no definition for uterine rupture.¹⁴⁸

Any induction method. The risk of uterine rupture associated with a TOL and IOL using any method was assessed using 14 fair quality studies, involving 12,659 women.^{11, 60, 96, 121, 125, 150, 154, 156–158, 160, 164, 167, 169} Combining these data results in a risk of 1.2 percent (95 percent CI: 0.9 to 1.6 percent) as shown below, with one small study reporting no uterine ruptures (0 percent) (Figure 18).

Figure 18

Risk of uterine rupture with any induction method. *95% confidence intervals are exact Abbreviation: GA= gestational age

Unfortunately, studies that reported the proportion of women with induced or augmented labors reported other maternal outcomes too infrequently to be meaningfully assessed.

Amnioinfusion. One small cohort study of women with prior cesarean delivery compared the rate of uterine rupture, defined as “full thickness separation of scar requiring operative intervention”, in women who underwent amnioinfusion during labor and those who did not.¹³⁸ However, exploration of the uterus was only undertaken among those with VBAC. The total number of women studied was 1,436, with 122 having received amnioinfusion. The rate of rupture was 0.8 percent in the amnioinfusion group and 1.1 percent in the group without amnioinfusion. There was no statistically significant difference in the rate of rupture between the groups, RR 0.72 (0.10 to 5.39). A smaller study reported no ruptures, but the definition and ascertainment of rupture was unclear.²¹²

Summary and strength of the evidence on effect of induction of labor on uterine rupture. The strength of evidence on the risk of uterine rupture with pharmacologic IOL methods was low due to lack of precision in estimates and inconsistency in findings. The overall risk of rupture with any IOL method at term was 1.5 percent and 1.0 percent when any GA is considered. Among women with GA greater than 40 weeks, the rate was highest at 3.2 percent. Evaluation of the evidence on specific methods of IOL reveal that the lowest rate occurs with oxytocin at 1.1 percent, then PGE₂ at 2 percent, and the highest rate with misoprostol at 6 percent. These findings should be interpreted with caution as there was imprecision and inconsistency in the results among these studies. The risk of uterine rupture with mechanical methods of IOL is understudied. Other harms were inadequately reported to make conclusions. Relative to women with spontaneous labor, there was no increase in risk of rupture among those induced at term. However, the available evidence on women with induced labor after 40 weeks GA indicates an increased risk compared with spontaneous labor (risk difference 1.8 percent; 95 percent CI: 0.1 to 3.5 percent). The NNH in this group is 56 (for every 56 women greater than 40 weeks GA with IOL during a TOL, one additional rupture will occur compared with having spontaneous labor).

Individual factors associated with uterine rupture. The impact of individual factors on uterine rupture can overlap and interact with each other such that a factor found to have statistically significant influence may no longer be significant when other key factors are taken into account. Studies that evaluate these factors in concert, using regression analyses for example, can provide a higher level of evidence on the residual influence of individual factors. Table 10 presents odds ratios reported by 11 good or fair quality studies that examined the relationship between individual factors and uterine rupture using the anatomic definition for uterine rupture.^{60, 80, 119, 150, 213–219} Individual factors associated with uterine rupture were grouped into four general categories: 1) demographic, 2) past obstetric factors, and 3) current obstetric factors. Nonclinical factors such as hospital type, VBAC volume, hospital delivery volume, maternal substance use or maternal medical conditions were only presented in studies that did not use the anatomical definition for uterine rupture or did not present a definition of uterine rupture and are therefore left off the table.

Table 10

The odds of uterine rupture after cesarean delivery by factor.

As shown, prior vaginal delivery and prior VBAC consistently appear to significantly reduce the risk of uterine rupture with odds ratios of 0.26 to 0.62 for prior vaginal delivery and 0.52 for prior VBAC. Inter-delivery interval less than 18 to 24 months and single layer closure appear to increase risk of uterine rupture with odds ratios of 2.05 to 2.65, and 3.95 to 4.33, respectively. However, caution should be used in interpreting the finding for layers of closure as this is based upon one study and the same study reports an increased risk for uterine rupture among preterm births, which is contrary to the reports of TOL and ERCD cohort studies.

Use of imaging to predict uterine rupture. The evidence on the use of various imaging modalities such as ultrasound, X-ray pelvimetry and endoscopy to predict uterine rupture risk is limited and consists of four fair quality cohort studies and six poor quality case series.^{82, 83, 222–229} The best evidence comes from three fair quality prospective cohort studies that measured full thickness lower uterine segment using ultrasound measurements between 35 and 38 6/7 weeks gestation.^{82, 83, 223} The first study conducted in France from 1989 to 1994⁸³ provides the most robust study design of the three to test the value of ultrasound as neither women nor their clinicians were informed of the results prior to delivery, thus removing the potential bias that could result from clinicians directing their patients towards or away from VBAC due to imaging results. In this study, women underwent ultrasound measurements of the lower uterine segment using a standardized method between 36 and 38 weeks and patients were divided into four groups according to full lower uterine segment thickness greater than 4.5 millimeters (mm), 3.6 to 4.5mm, 2.6 to 3.5mm and 1.6 to 2.5mm. There was a significant association between uterine wall thinning and uterine scar defects, defined as uterine rupture or dehiscence. Overall, the relative risk of a defect was 20.1 percent (95 percent CI: 8.3 to 48.9 percent) for lower uterine segment less than or equal to 3.5mm and 6.3 percent (95 percent CI: 2.8 to 13.9 percent) for measurements less than or equal to 2.5mm. While measurements greater than 4.5mm had a negative predictive value of 100 percent given no cases of defect in this group, the positive predictive value for a thickness less than 3.5mm was not good at 11.8 percent (negative predictive value 99.3 percent). Of note, this population included women with multiple prior cesarean deliveries and there was no clear association between number of prior cesareans and uterine thickness measurements. This study did not adjust for other factors that might also contribute to risk such as direction of prior cesarean scar, estimated fetal weight, co-morbidities, indication for prior cesarean delivery, history of prior vaginal delivery etc. While the two additional studies that follow, Rozenberg 1999⁸² and Bujold 2009,²²³ have higher risk of bias as clinicians were informed of the imaging results prior to delivery, both support the association between uterine thickness and uterine rupture (Table 11).

Table 11

Lower uterine segment thickness and uterine defect.

Strength of evidence for individual factors as predictors of uterine rupture. Studies of individual factors that may increase or decrease a woman’s risk of uterine rupture are largely exploratory, as few factors have been confirmed in prospective studies as suggested by Simon and Altman, 1994.¹⁹³ There is cross validating evidence to suggest that women with prior vaginal delivery have lower risk for uterine rupture and women undergoing IOL have higher risk of uterine rupture compared with spontaneously laboring women. Similarly, evidence from IOL studies suggest that women who are postdates may have a higher risk of uterine rupture. The evidence on the role of imaging to predict uterine rupture is low due to limited studies with high-risk for bias; however, the existing data suggest that there may be value to ultrasound measurements of uterine thickness for women with prior cesarean delivery.

Predictive models for uterine rupture. Because uterine rupture is such an important consideration for women with prior cesarean delivery, several investigators have attempted to build predictive tools that would combine the individual predictors to estimate a woman’s risk for uterine rupture. Four studies attempted to develop predictive models for uterine rupture (see Appendix P for study details).^{104, 215, 219, 230} Only two tested their predictive model in a validation group.^{104, 215} No study was able to produce a reliable and robust model to predict uterine rupture.

Signs of uterine rupture and management to reduce uterine rupture related mortality. Given the serious potential consequences for morbidity and mortality given uterine rupture, it would be ideal to understand signs and symptoms of uterine rupture and what, if any interventions might reduce the likelihood of morbidity or mortality in the event that a uterine rupture does occur. As stated in a prior VBAC evidence report, and echoed in studies contained in this report, there is no single sign for the occurrence of uterine rupture; however, fetal heart tracing abnormalities, particularly fetal bradycardia (reported in 33 to 100 percent of all studies) is the most frequently reported sign of uterine rupture.^{62, 206–208} Other signs reported in uterine rupture studies in descending order are maternal vaginal bleeding, maternal pain, and uterine contraction disturbances. ⁶²

The next important question is whether there are any management or system factors that might reduce the likelihood for an adverse outcome in the event of uterine rupture. Two fair quality case series^{231, 232} have specifically studied whether time from signal of uterine rupture to delivery predicts perinatal outcomes. Leung et al were the first to perform an exploratory analysis to study risk factors for poor neonatal and maternal outcome; particularly fetal heart rate (FHR) and uterine contraction patterns.²³³ They identified 106 cases of symptomatic uterine rupture from 11,179 TOLs in women with prior cesarean delivery at Los Angeles County University of Southern California Women’s Hospital, from which they were able to review 99 records. 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. Looking for premonitory signs of uterine rupture, they found that prolonged decelerations occurred in 17/41 (41.5 percent) patients with extrusion and 15/58 (25.9 percent) without and that no patient who had prolonged deceleration only as their sign of rupture had significant clinical morbidity when delivery occurred within 17 minutes of the onset of deceleration. Four of six infants (67 percent) who died due to uterine rupture presented to labor and delivery in “fetal distress” and two of six occurred in women undergoing TOL. It is important to note that 57 percent of uterine ruptures with fetal extrusions occurred in women with two or more prior cesarean deliveries; 21 percent in classical or vertical incisions, 21 percent in unknown incisional type, and 58 percent in LTCD.

A second case series of 23 uterine ruptures out of Canada found no relation between time from FHR deceleration and infant outcome.²³¹ As above, the study was conducted in a tertiary care hospital with in-house anesthesia and obstetrics. Fetal heart rate abnormalities—which included tachycardia and late, variable, or prolonged (not defined) decelerations—were the initial sign of uterine rupture in 20/23 (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 delivery), and three cases of impaired motor development diagnosed as hypoxic-ischemic encephalopathy (HIE), 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.

Summary and strength of the evidence on uterine rupture. Overall, the literature relating to response time between premonitory signs of uterine rupture and perinatal mortality are insufficient. This is due to study designs that are more prone to bias, inconsistent findings, imprecision and difficulty accounting for time among women who presented with concerning fetal tracings (e.g. whether the patient had concerning tracing at arrival to the hospital). However, there is suggestion that fetal bradycardia is an ominous sign for fetal extrusion, which is associated with poor perinatal outcomes, and prompt delivery in this setting is warranted.

Hysterectomy

Overall there were 16 cohort studies reporting hysterectomy as a complication of ERCD, VBAC, and RCD after a TOL;^{10, 77, 78, 80, 81, 93, 100, 108, 137, 164, 201, 204, 228, 234–236} eight provided information comparing risks for hysterectomy between TOL and ERCD (Table 12).^{10, 77, 81, 93, 108, 201, 204, 234} As shown in Figures 19 and 20, while the occurrence of hysterectomy was higher for ERCD at 0.28 percent (95 percent CI: 0.12 percent to 0.67 percent) compared to 0.17 percent (95 percent CI: 0.12 to 0.26 percent) for TOL but the difference was not statistically significant. Three of the eight studies focused exclusively on women delivering at term.^{93, 108, 204} Among these studies, the combined risk of hysterectomy was 0.14 percent (95 percent CI: 0.08 to 0.22 percent) in the TOL group and 0.16 percent (95 percent CI: 0.07 to 0.36 percent) in the ERCD group. The risk was not significantly different between the two groups (p=0.672). Among the five studies including women delivering at any gestational age, the combined risk of hysterectomy was 0.22 percent (95 percent CI: 0.13 to 0.38 percent) for TOL and 0.43 percent (95 percent CI: 0.11 to 0.17 percent) for ERCD. Compared with ERCD, TOL had a significantly lower risk of hysterectomy (RR, 0.40; 95 percent CI: 0.18 to 0.92, p=0.03). Using 0.43 percent as the baseline risk for ERCD, the calculated risk difference was −0.26 percent (95 percent CI: −0.35 to −0.04 percent), translating to 2.6 fewer hysterectomies per 1,000 for TOL.

Table 12

Hysterectomy for trial of labor versus elective repeat cesarean delivery.

Figure 19

Hysterectomy occurrence for trial of labor.

Figure 20

Hysterectomy occurrence for elective repeat cesarean delivery.

There was significant heterogeneity among the studies. In particular, Phelan and Eglinton appeared to be outliers for ERCD among term studies and because of its size, Wen is exerting significant effect on all analyses and particularly for ERCD; this study has a low occurrence of hysterectomy. Studies were explored for factors that might explain the observed heterogeneity. Considered factors included setting (university versus community), gestational age, number, and direction of scar, and no clear pattern was observed across the eight studies.

Individual studies looked for factors that may affect the occurrence of hysterectomy among women with prior cesarean. In particular, several publications from the MFMU study, report interesting potential contributors. Grobman et al examined the influence of induction of labor and reported that induction was associated with almost a 4-fold increase in the occurrence of hysterectomy (odds ratio 3.92; 95 percent CI: 1.10 to 13.9) among women without a history of prior vaginal delivery. Mercer et al evaluated the risk of hysterectomy with increasing number of VBACs found a non-significant decrease in the rate of hysterectomy from 0.23 percent with no prior VBAC to 0.016 percent in subjects with two or more previous VBACs (p=0.15).¹⁹² Perhaps the most intriguing factors reported from this group is in regards to the influence of multiple prior cesarean deliveries and prior vaginal delivery. In subjects undergoing multiple cesarean deliveries, there was a decrease in hysterectomy rate between the first and second cesarean delivery, followed by an increase from 0.42 percent after two cesarean deliveries to 9.0 percent after six or more previous cesarean deliveries.²³⁶ Only one study evaluating this cohort compared hysterectomy rate among VBAC after one cesarean delivery, VBAC after greater than one cesarean delivery, and ERCD. This study found the lowest proportion of hysterectomy among those with VBAC after a single cesarean delivery (0.2 percent), intermediate levels after ERCD (0.4 percent), and the highest among those with VBAC after multiple previous cesarean deliveries (0.6 percent).⁸⁰ These two studies suggest multiple cesarean deliveries increase risk for hysterectomy at the time of delivery; VBAC may be protective against hysterectomy after multiple cesarean deliveries. However, this comparison is limited by a lack of information about the actual number of previous cesarean deliveries and TOL in the latter article, and therefore a dose–dependent effect of increasing number of cesarean deliveries with TOL and VBAC cannot be determined. Furthermore, the inclusion criteria for ERCD limited its applicability, as only low risk ERCD candidates were included. Using data from a California state database, Gregory et al found that underlying medical and obstetrical risk may increase a woman’s chance for hysterectomy reporting a two-fold increase in hysterectomy for ERCD compared with TOL among women with high-risk pregnancies (0.41 versus 0.22 percent).⁹³

Summary of hysterectomy. Hysterectomy is rare with either ERCD or TOL, occurring in less than 3 percent of deliveries for women with a prior cesarean delivery. There was no significant difference in the occurrence of hysterectomy based upon route of delivery overall or among women delivering at term; however there was a lower occurrence of hysterectomy among women undergoing a TOL among studies enrolling women of any gestational age.

Transfusion/Hemorrhage

Transfusion. Nine cohort studies of good or fair quality totaling 401,307 patients evaluated the occurrence of transfusion between ERCD and TOL^{10, 76, 77, 93, 95, 97, 108, 204, 234} (see Appendix Q for a detailed description of studies). As shown in Figure 21, the occurrence of transfusion was not significantly different between TOL and ERCD (0.9 versus 1.2 percent, this translates to nine versus 12 per 1,000). There was significant heterogeneity among the studies however with studies varying widely on the frequency of transfusion, with transfusion rates ranging from 0.5 to 4.3 percent for TOL and 0.1 to 5.5 percent for ERCD. Studies were also conflicting on which group had the highest rate of transfusion with five studies reporting greater transfusions with ERCD,^{10, 76, 77, 95, 234} and four reporting greater transfusions in TOL.^{93, 97, 108, 204} The studies were examined for potential factors that may explain the difference. Factors that were reported among studies included gestational age, accuracy of group assignment, maternal conditions particularly underlying high-risk medical conditions, hospital setting, and the influence of prior vaginal delivery.

Figure 21

Transfusion rates for trial of labor versus elective repeat cesarean among all studies.

Term studies. Four, of the nine studies focused exclusively on women delivering at term (Figure 22).^{93, 97, 108, 204} All four term studies reported higher rates of transfusion among women who had a TOL, with two reaching statistical significance.^{97, 204} The combined risk of transfusion was 0.7 percent (95 percent CI: 0.2 to 2.2 percent) for TOL and 0.5 percent (95 percent CI: 0.2 to 1.3 percent) for ERCD, translating to seven per 1,000 and five transfusions per 1,000, respectively. When these studies were combined there was a significantly increased risk of transfusion for TOL compared with ERCD (RR, 1.30; 95 percent CI: 1.15–1.47; p<0.001). Using 0.5 percent as the baseline risk for ERCD, the calculated risk difference was 0.14 percent (95 percent CI: 0.07 to 0.22 percent), which is equivalent to 1.4 more transfusion per 1,000 for TOL.

Figure 22

Transfusion rates for trial of labor versus elective repeat cesarean delivery among term studies.

When looking at the five studies of women delivering at any gestational age, the combined risk of transfusion was 1.2 percent (95 percent CI: 0.7 to 2.3 percent) for TOL and 2.4 percent (95 percent CI: 1.3 to 4.3 percent) for ERCD, translating to 12 per 1,000 and 24 transfusions per 1,000 respectively. There was a significantly reduced risk of transfusion for TOL compared with ERCD (RR, 0.48; 95 percent CI: 0.30 to 0.79; p=0.003). Using 2.4 percent as the baseline risk for ERCD, the calculated risk difference was −1.26 percent (95 percent CI: −1.71 to −0.52 percent), which is equivalent to 12.6 fewer transfusions per 1,000 for TOL.

Group assignment. The disparities among the pooled analyses raise questions about whether other factors may be playing a role. Two studies provided separate results for VBAC, ERCD, and cesarean after TOL.^{95, 234} In both studies, VBAC had the lowest rate of transfusion, followed by cesarean after TOL, while ERCD had the highest rate of transfusion. This suggests the influence of term gestational ages may be related to increased transfusions with cesarean after TOL; however, the term gestation delivery studies did not report these outcomes.

Maternal factors. The pooled risk is strongly influenced by the largest study, which included all levels of care as well as relied on ICD-9 codes for data gathering. This study found opposing results to the general trends seen with the other studies in this group, and is likely strongly contributing to the effect seem with the pooled data.¹⁰⁸ Two studies examined the affect of medical conditions upon transfusion rates.^{93, 204} Both studies found an association between high-risk pregnancies with medical conditions and increased transfusion. Gregory found that among high-risk patients, women who delivered by ERCD had more transfusions than those who had a TOL (0.92 versus 0.78 percent, p=NS) and that the overall rate of transfusions was greater among higher risk pregnancies compared with the low-risk group (0.46 versus 0.33 percent, p=NS).⁹³ Similarly, using data from the MFMU cohort, Spong et al divided the type of cesarean delivery into ERCD and indicated repeat cesarean (IRCD) with and without labor, and found the highest proportion of transfusion with IRCD in the absence of labor, again suggesting an influence of maternal co-morbid conditions contributing to increased transfusion risk.²⁰⁴ An analysis of the MFMU cohort focusing only on low-risk ERCD found a statistically significant increase in transfusion with TOL compared with ERCD prior to labor (one versus 1.7 percent, p<0.001), further supporting the idea that maternal co-morbid conditions influence the risk of transfusion.²²¹ In this same cohort, however, increasing maternal BMI and morbid obesity were not found to be associated with increased risk of transfusion in the MFMU cohort.⁷⁸

Hospital setting. One study evaluated the impact of hospital delivery volume on the risk of transfusion.¹⁰⁸ In low volume hospitals (less than 500 deliveries per year) the odds of transfusion with TOL was 1.39 (1.02 to 1.88) compared with ERCD. In high volume centers (greater than 500 deliveries per year) the odds ratio was 1.66 (1.32 to 2.08) for transfusion with TOL compared with ERCD.¹⁰⁸ This study differs in its findings from other studies as described previously; confounding factors that may contribute to this difference is that this review of a birth registry was based on ICD-9 codes; also, unlike the other studies, this study does report only on Canadian hospitals, and includes multiple different levels of hospital acuity.

Influence of prior deliveries. Within the MFMU cohort, multiple studies evaluated the impact of previous deliveries on the risk of transfusion.^{80, 181, 192, 236} Subjects with only one prior cesarean delivery undergoing TOL had a rate comparable with ERCD (1.6 versus 1.5 percent), while those with multiple prior cesareans undergoing TOL had a higher transfusion rate compared with ERCD (3.2 versus 1.5 percent).⁸⁰ This study did not distinguish those with ERCD undergoing multiple cesareans versus first RCD. However, increasing number of cesareans was found to be a risk factor for transfusion.²³⁶ The MFMU cohort had a decreasing rate of transfusion with increasing number of previous VBACs, implying a protective role for prior vaginal delivery against transfusion (zero prior VBAC: 1.89 percent; one prior VBAC: 0.24 percent; two or more prior VBAC: 0.99 percent; p=0.002).¹⁹² One study of this cohort found no difference in women with a prior vaginal delivery when comparing those who were induced with those who went into spontaneous labor (odds ratio 1.13; 95 percent CI: 0.66 to 1.95). However, in those women without a previous vaginal delivery, there was a higher rate of transfusion in the induced group compared with the spontaneous labor group, suggesting a role of induction as a risk for transfusion in women with no prior vaginal delivery only (odds ratio 1.65; 95 percent CI: 1.10 to 2.48).¹⁵⁶

Hemorrhage. Six fair quality cohort studies report on the occurrence of hemorrhage for TOL versus ERCD.^{76, 93, 95, 100, 203, 234} Among the six studies the rates of hemorrhage with ERCD ranged from 0.3 percent to as high as 29 percent. In general, studies reported increased occurrence of hemorrhage associated with ERCD compared to TOL. However, studies were inconsistent regarding the definition of hemorrhage. Several studies did not report the process by which this outcome was measured; studies that quantified blood loss used different amounts of blood loss to qualify as hemorrhage. The highest rate of hemorrhage was from a study that defined hemorrhage as greater than 500mL blood loss.⁷⁶ Further complicating evaluation of these data is the known difficulty in recording blood loss. Multiple studies demonstrate physician perception of blood loss differs from actual blood loss, impeding accurate recording of this measure.^{237, 238} All studies found a trend toward increased blood loss with ERCD. However, none of the studies found a statistically significant difference between hemorrhage rates for TOL and ERCD. Because these studies did not define hemorrhage similarly, these data were not pooled for analysis. Interesting findings from individual studies are discussed below.

Term studies. Only one of the six studies,⁹³ provided data regarding hemorrhage specifically in term pregnancies. In this study, low-risk patients were separated from high-risk patients based on antenatal conditions. The low-risk group had a lower rate of hemorrhage with TOL compared with ERCD (2.36 versus 6.82 percent, p=NS); however in the high-risk group, there was an increase in hemorrhage with TOL compared with ERCD (3.26 versus 1.57 percent). In the one study that reported both, the rates of transfusion are in direct opposition to the rates of hemorrhage for every subset, such that in groups with higher rates of hemorrhage, there were overall fewer transfusions given.⁹³ This was a study that used an administrative database and demonstrates the difficulty in reporting and interpreting a subjective term such as hemorrhage.

Group assignment. Three studies separated the TOL data by VBAC and cesarean after a TOL.^{95, 203, 234} Rates of hemorrhage for VBAC ranged from 0.3 to 6 percent, compared with cesarean after TOL which ranged from 0.86 to 14.8 percent. In three of these studies, cesarean after a TOL had the highest rate of hemorrhage compared to VBAC and ERCD. One study found a statistically significant increase in hemorrhage with ERCD compared to cesarean after a TOL and VBAC (0.91 versus 0.64 versus 0.81 percent, p<0.001).⁹⁵ This study did not define hemorrhage; in addition, this university-based Israeli study focused on a grand multiparous subgroup, possibly confounding the results given the global increased risk of hemorrhage. One database study from Norway described the incidence of postpartum hemorrhage. Though it did not specifically evaluate women undergoing TOL or ERCD, it did find 333 cases of hemorrhage in women with a history of previous cesarean delivery. This study found an increased risk of hemorrhage with both vaginal delivery (odds ratio 1.63; 95 percent CI: 1.34 to 1.98) and emergency cesarean (1.41; 95 percent CI: 1.12 to 1.78) in women with a previous history of cesarean compared to women without a history of cesarean.²³⁹ This study found no difference in hemorrhage between these two groups for pre-labor cesarean; however, for women with prior cesarean, pre-labor cesarean carried a 28 percent higher risk of hemorrhage compared to spontaneous labor delivery.²³⁹

Summary and strength of evidence on transfusion/hemorrhage. Overall, evidence regarding the rate of hemorrhage for TOL and ERCD is low due to inconsistency in definitions and subjectivity in measurement. In general, studies reported increased occurrence of hemorrhage associated with ERCD compared to TOL. There was only one study that provided data for term populations. This study suggested that medical complications may modify the effect of route of delivery upon hemorrhage with low-risk patients similarly having higher hemorrhage rates for ERCD but with high-risk patients experiencing higher rates of hemorrhage for TOL. Further studies are needed to understand the true relationship.

Infection

Twenty-two studies of good or fair quality evaluated infectious morbidity (e.g., fever, infection, endometritis, or chorioamnionitis) in TOL compared with ERCD (see Appendix R for study details).^{10, 76–79, 81, 89, 95, 97, 108, 114, 201–205, 221, 234, 240–243} Ten studies ^{10, 89, 95, 97, 108, 202–204, 234, 240} report on infection in some manner. As shown in Figure 23, there was no significant difference in infection between TOL and ERCD. ). The I² statistic for heterogeneity was 89 percent. However, the confidence in the magnitude and direction of the estimates from this body of literature is low due to inconsistencies in definitions, indirect evidence, and high risk of bias. Details in the specific infections reported in these studies are described below.

Figure 23

Rates of infection for trial of labor versus elective repeat cesarean delivery among all studies.

Endometritis. Six studies of good or fair quality compared endometritis in TOL with ERCD.^{89, 95, 203, 221, 240, 241} Overall, rates of endometritis were higher in TOL when compared with ERCD. Rates of endometritis in TOL ranged from 0.8 to 30 percent. The rates of endometritis in ERCD ranged from 1.2 to 18 percent. The upper ranges of these figures both derive from the same study focusing on morbidly obese patients, suggesting an influence of obesity on the risk of infection.²⁴⁰

The MFMU cohort evaluated endometritis in TOL compared with ERCD (Appendix R).²²¹ A separate report on this same cohort specifically evaluated BMI and the risk of endometritis in morbidly obese patients.⁷⁸ There was a statistically significant increase in the rate of endometritis with increasing BMI in patients undergoing a TOL. Additionally, there was a statistically significant increased odds of endometritis for TOL compared with ERCD in morbidly obese (BMI greater than 40) subjects. (odds ratio 2.4; 95 percent CI: 1.7 to 3.5).⁷⁸ This again suggests maternal weight influences the risk of endometritis in this population.

In studies where outcomes of TOL were evaluated separately, cesarean after labor patients consistently had a higher rate of endometritis than did VBAC or ERCD patients.^{89, 95, 203} This suggests the increased rates of endometritis seen with TOL may be more associated with cesarean delivery than with the labor itself.

Chorioamnionitis. In total, two cohort studies compared chorioamnionitis rates for TOL and ERCD in deliveries.^{89, 95} Both found a higher rate of chorioamnionitis in the TOL group. Based on this, there is evidence that chorioamnionitis is associated with TOL, regardless of ultimate mode of delivery.

Wound infection. Wound infection was evaluated in three cohort studies.^{10, 203, 240} Two of the three found ERCD to be associated with a higher rate of wound infection, ^{10, 203} while one found a higher rate with TOL. This study was limited to women over 300lbs at delivery, which complicates any direct comparison of this data with the others.²⁴⁰ None of these studies reached statistical significance with their findings.

Fever. Ten good or fair quality cohort studies compared maternal fever between TOL and ERCD.^{10, 76, 77, 79, 81, 89, 95, 201, 202, 205} As shown in Figure 24, the combined absolute risk for any fever with TOL was 6.5 percent (95 percent CI: 4.4 to 9.3 percent) which translates to 65 per 1,000 (95 percent CI: 44 to 93 per 1,000) and for ERCD was 7.2 percent (95 percent CI: 2.5 to 18.9 percent) which translates to 72 per 1,000 (95 percent CI: 25 to 189 per 1,000). Compared with ERCD, TOL demonstrated a significant decrease in the risk of fever (RR, 0.63; 95 percent CI: 0.43 to 0.91; p=0.013). Using 7.2 percent as the baseline risk for ERCD, the calculated risk difference was −2.7 percent (95 percent CI: −4.10 to −0.68 percent), which is equivalent to 27 fewer fevers per 1,000 from the TOL There was considerable heterogeneity among studies (I² = 92.7 percent), which demonstrates the variability in the definition of the term “fever” among studies. Only one study in this pooled analysis defined fever by an absolute number (greater than 38 degrees Celsius); it is interesting to note this study did serve as the most apparent outlier in the pooled analysis.⁷⁹ When this study was excluded from analysis, the combined risk of fever for ERCD group was 11.0 percent and TOL still had a significantly lower risk of fever (RR, 0.58; 95 percent CI: 0.40 to 0.82; p=0.002). Using 11.0 percent as the baseline risk for ERCD, the calculated risk difference was −4.68 percent (95 percent CI: −6.58 to −1.95 percent), which is equivalent to 47 fewer fevers per 1,000 for TOL. The I² statistic for heterogeneity remained 92.7 percent, demonstrating that this outlier did not explain the heterogeneity seen among studies.

Figure 24

Rates of fever for trial of labor versus elective repeat cesarean delivery among all studies.

Four cohort studies evaluated maternal fever stratified by outcome of TOL.^{81, 89, 95, 201} Three of the four studies found increased rates of fever in cesarean after a TOL compared with VBAC.^{81, 89, 201} This suggests a higher febrile morbidity associated with cesarean delivery after labor compared with VBAC. However, in all but one of these studies ERCD continued to have the highest rate of febrile morbidity, suggesting surgery as a risk factor for maternal infectious morbidity. This would imply the rate of fever seen with TOL may be influenced by the higher risk of fever with RCD after a TOL and not be associated with VBAC. The one outlier study in this group found a rate of 32 percent for cesarean after a TOL compared with 18 percent for ERCD. However, the population of this study was limited to subjects with two prior cesarean deliveries. Increased complications with second cesarean after labor may have influenced the higher rates seen with ERCD.⁸¹ Within women who underwent a TOL, there appears to be an increased risk for febrile morbidity with operative delivery.

Summary and strength of evidence on infection. Overall, the evidence regarding rate of both infection (all definitions) and fever for women with a prior cesarean delivery is low in strength with inconsistent definitions, high-risk of bias and indirect evidence. The high heterogeneity in these studies demonstrates the variability in the definition of the term “fever” among studies. Overall there was no significant difference in infection between TOL and ERCD.

Surgical injury

In good or fair quality cohort studies, surgical injury was defined differently between studies and variably reported on. Seven studies compared surgical injury between TOL and ERCD.^{10, 78, 80, 97, 204, 205, 236} Four of these are from the same cohort of patients (MFMU); however they report differently on surgical injury rates and therefore are presented separately (Table 13).^{78, 80, 204, 236}

Table 13

Surgical injury rates for trial of labor versus elective repeat cesarean delivery.

Overall, these were found to be rare events. None of the stratifications from the MFMU studies found a significant difference between ERCD and TOL for rate of surgical injury.^{78, 80, 204, 236} When a specific injury was evaluated, such as bladder, there did tend to be an increased rate of injury with a cesarean after a TOL. However, this trend was not found to be statistically significant.²⁰⁵

One case-control study evaluating women with bladder injury at the time of cesarean delivery found attempted VBAC to be proportionately higher in the cases compared with controls (64 versus 22 percent, p<0.1) This study found no difference in type of uterine incision with risk of bladder injury.²⁴⁴

One study of 3,164 women evaluated the impact of skin incision direction on the risk of bowel or bladder injury. This retrospective study found an increased odds ratio of bladder injury with a midline sub-umbilical incision compared with a Pfannenstiel (6.7; 95 percent CI: 2.6 to 16.5). The RR for bowel injury was also increased with vertical incision at five and a half fold risk. After multivariate analysis, vertical abdominal incision remained a significant risk factor for bladder injury at the time of cesarean. There was a trend toward increased injury with increasing numbers of cesarean deliveries, though this was not statistically significant.²⁴⁵

Summary of surgical injury. Rate of surgical injury may be increased with TOL but definitive studies are lacking. Vertical skin incision increases risk of surgical injury to the bladder.

Hospital Stay

Hospital stay was reported as length of total stay in days. A total of eight cohort studies examined length of stay data in the U.S., comparing ERCD and TOL (Appendix S).^{77, 79, 81, 93, 97, 201, 202, 234} All studies were affiliated with teaching institutions. In general, as expected, ERCD had a longer length of hospital stay compared with TOL. The large MFMU cohort studies did not report length of stay data comparing ERCD with TOL. One study did evaluate the risk of extended stay, defined as greater than 4 days, with ERCD compared with TOL in morbidly obese patients; this study found increased length of stay with ERCD (odds ratio 1.2; 95 percent CI: 1.1 to 1.4).⁷⁸

Pooled analysis of any gestational age studies. For any GA cohorts, the pooled analysis using a random effects model demonstrated that the mean length of stay for TOL was 2.55 days (95 percent CI: 2.34 to 2.76 days). Pooled mean length of stay for ERCD was 3.92 days (95 percent CI: 3.56 to 4.29 days). There was significant heterogeneity among studies (Q=221.06, p<0.001) and the I² statistic for heterogeneity was 98.2 percent (between-study heterogeneity accounts for 98.2 percent of the total heterogeneity) (Figure 25).

Figure 25

Length of stay for trial of labor versus elective repeat cesarean delivery among any gestational age studies.

Summary of hospital stay. Elective repeat cesarean delivery is associated with a longer hospital stay compared with TOL.

Pelvic Floor

One study evaluated the effect of VBAC on perineal trauma in deliveries over 36 weeks GA. After controlling for age, parity, and episiotomy, those women who had a VBAC as their second delivery had an odds ratio of 5.46 (95 percent CI: 3.69 to 8.08) of severe perineal trauma compared with women with a previous vaginal delivery. In contrast, the odds ratio for a primiparous woman was 4.08 (95 percent CI: 3.16 to 5.28).²⁴⁶ A second study noted a 34 percent prevalence of third or fourth degree perineal laceration following episiotomy in women undergoing VBAC; however no comparisons were made to RCD after a TOL.¹³⁷ No studies evaluated risk of urinary or fecal incontinence following ERCD versus TOL.

Deep Vein Thrombosis

Three studies evaluated thromboembolic disease between ERCD and TOL.^{192, 203, 221} A multicenter study evaluating maternal risk and thromboembolic disease found the lowest rate of embolic disease with TOL after one prior cesarean delivery compared with subjects undergoing either ERCD or VBAC after multiple cesarean deliveries (0.04 percent versus 0.1 and 0.1 percent).²²¹

Special Considerations

Effect of hospital setting. In total, eight studies reported outcomes in both community and academic centers.^{10, 61, 77, 93, 97, 98, 108, 205} Three of these studies reported on the same cohort of patients.^{61, 98, 205} Of these eight studies, only one specifically described the community data separately.¹⁰⁸ As described previously, this study evaluated maternal risk with ERCD, TOL, VBAC, and RCD using the Canadian birth registry. This evaluation focused specifically on the risks of each mode of delivery in low volume (less than 500 deliveries per year) and high volume (greater than 500 deliveries per year) medical centers. This study demonstrated an increased odds of short-term complications in low volume maternity wards. Specifically, the odds ratio of a particular outcome associated with TOL when compared with ERCD were higher for death (odds ratio 2.68; 95 percent CI: 0.16 to 45.5 versus odds ratio 0.16; 95 percent CI: 0.02 to 1.29) and uterine rupture (odds ratio 4.02; 95 percent CI: 2.48 to 6.51 versus odds ratio 2.30; 95 percent CI: 2.04 to 2.59) in low volume maternity wards.¹⁰⁸

Abnormal placentation. Prior cesarean delivery is a risk factor for abnormal placentation in future pregnancies. As the number of cesarean deliveries continues to rise, the incidence of abnormal placentation, which includes placenta previa, accreta, increta, and percreta, is anticipated to increase as well. Abnormal placentation has been associated with both maternal and neonatal morbidity including need for antepartum hospitalization, preterm delivery, emergent cesarean delivery, hysterectomy, blood transfusion, surgical injury, intensive care unit (ICU) stay, and fetal and maternal death. In order to effectively counsel women about their risk of complications due to the placenta in future pregnancies, it is essential to have a clear understanding of the incidence of these potentially life-threatening complications in women with prior cesarean delivery.

Incidence of abnormal placentation following cesarean section. Eighty-two full text articles were reviewed to evaluate the evidence regarding the incidence and outcomes of pregnancies complicated by abnormal placentation, including abruption, placenta previa, and placenta accreta, following prior cesarean delivery. Nineteen articles met inclusion criteria and consisted of eight good or fair quality cohort studies,^{11, 220, 236, 247–251} seven fair quality case-control studies,^{176, 235, 252–256} and four good or fair quality case series.^257–260 Individual studies provided evidence for one or more of the separate topics of abruption, previa, and accreta, respectively. The studies that made up the body of evidence for each subset of abnormal placentation are listed below.

Abruption. Six fair quality studies, five cohort studies^{11, 247, 249–251} and one case-control study²³⁵ examined abruption following a prior cesarean delivery. Studies were inconsistent in their definition for placental abruption with three studies relying on ICD-9 coding^{11, 247, 250} and no definition offered by the other three.^{235, 249, 251}

The overall incidence of placental abruption with any prior cesarean delivery was 1.2 to 1.5 percent.^{250, 251} For women with one prior cesarean delivery, the odds ratio for abruption was 1.0 to 1.3 and only one of four studies reached statistical significance. The majority of studies did not find an increased incidence of abruption in women with increasing numbers of prior cesarean delivery (Table 14).

Table 14

Abruption based on number of prior cesarean deliveries.

Women with abruption were more likely to require blood transfusion, but the incidence did not increase with increasing number of cesarean delivery. Of women without a prior cesarean delivery, 14.3 percent required blood transfusion compared with 14.1 percent in women with one or more prior cesarean deliveries without abruption. Other maternal outcomes such as hemorrhage and hysterectomy were reported inconsistently such that meaningful analysis of an association with abruption and prior cesarean delivery was not possible (Table 15).²⁵¹

Table 15

Blood transfusion by number of prior cesarean deliveries.

Placenta previa. Eight good or fair quality cohort studies,^{11, 220, 236, 247–251} five fair quality case-control studies,^{235, 253, 255, 259, 261} and three good or fair case series^{257, 258, 260} provide the primary body of evidence regarding placenta previa following prior cesarean delivery. Only six studies provided information on the incidence of placenta previa following prior cesarean delivery (Table 16).^{11, 248, 251, 258, 259, 261} Again, studies differed in their definition for placenta previa with two studies using a previa grade scale and the remaining providing individual definitions.

Table 16

Overall incidence of placenta previa.

Women with a prior cesarean delivery had a statistically significant increased risk of placenta previa compared with women with no prior cesarean delivery (odds ratio 1.48 to 3.95, Table 17). The studies conflicted as to whether the risk increased with increasing cesarean deliveries. The incidence of previa with one prior cesarean delivery was 0.8 to 1.5 percent. Compared with women without a prior cesarean delivery, the odds ratio was 1.2 to 1.9. This was statistically significant in four of seven studies. In women with two prior cesarean deliveries, the incidence of previa was 1.1 to 2.0 percent with an odds ratio of 1.9 to 2.0; this was statistically significant in three of five studies. Two studies limited comparisons to women with one prior cesarean delivery versus multiple cesarean deliveries.^{248, 249} In these studies, no increased risk with additional cesarean deliveries was noted. These studies had limited sample size for women with multiple cesarean deliveries and previa (N=7 and N=20, respectively).^{248, 249} The six studies that specifically identified women with three or more cesarean deliveries all noted a statistically significant increased rate of previa with increasing cesarean deliveries, up to 3.7 percent for women with five or more prior cesarean deliveries.^{235, 236, 248, 250, 253, 258}

Table 17

Incidence of placenta previa by number of prior cesarean deliveries.

The pooled analysis using a random effects model demonstrated absolute risk of previa associated with any number of prior cesarean deliveries was 1.2 percent (95 percent CI: 0.8 to 1.5 percent), one prior cesarean delivery was 1 percent (95 percent CI: 0.6 to 1.3 percent), two prior cesarean deliveries was 1.7 percent (95 percent CI: 1.1 to 2.3 percent) two or more cesarean deliveries was 2.3 percent (95 percent CI: 1.1 to 3.4 percent), three or more cesarean deliveries was 2.8 percent (95 percent CI: 1.8 to 3.7 percent, Figure 26). This translates to any number of prior cesarean deliveries was 12 per 1,000 (95 percent CI: 8 to 15 per 1,000), one prior cesarean delivery was 10 per 1,000 (95 percent CI: 6 to 13 per 1,000), two prior cesarean deliveries was 17 per 1,000 (95 percent CI: 11 to 23 per 1,000) two or more cesarean deliveries was 23 percent (95 percent CI: 11 to 34 per 1,000), three or more cesarean deliveries was 28 percent (95 percent CI: 18 to 37 per 1,000).

Figure 26

Incidence of placenta previa by number of prior cesarean deliveries among any gestational age studies.

The incidence of hysterectomy increased in women with placenta previa depending on the number of prior cesarean deliveries (Table 18). Women with no prior cesarean delivery and previa required hysterectomy in 0.7 to 4 percent of cases compared with 50 to 67 percent in women with three or more prior cesarean deliveries. Women with prior cesarean delivery and previa were more likely to require a blood transfusion (odds ratio 15.9; 95 percent CI: 12.0 to 21.0). Two authors also reported composite major maternal morbidity in women with prior cesarean delivery and previa. One study evaluated severe postpartum hemorrhage, acute renal failure, ICU admission, mechanical ventilation, shock, disseminated intravascular coagulopathy, hysterectomy, other procedures to stop bleeding, and/or death.²⁵⁹ Thirty percent of women with prior cesarean delivery and previa experienced major maternal morbidity (odds ratio 3.1; 95 percent CI: 2.0 to 4.7). A second study combined transfusion, hysterectomy, operative injury, coagulopathy, venous thromboembolism, pulmonary edema, and/or death.²²⁰ There was a statistically significant increase in composite major maternal morbidity with increasing number of cesarean delivery from 15 percent with no prior cesarean delivery to 83 percent with three or more prior cesarean deliveries (odds ratio 33.6; 95 percent CI: 14.6 to 77.4).

Table 18

Incidence of hysterectomy in women with previa by number of prior cesarean deliveries.

Placenta accreta. Three good or fair quality cohort studies^{236, 249, 258}, one fair quality case-control study,²⁵⁶ and one fair quality case series study²⁶⁰ examined placenta accreta and prior cesarean delivery and the relationship between placenta previa and accreta (Table 19). Two fair quality studies evaluated placenta accreta and hysterectomy.^{252, 254} Studies varied widely on the definition of placenta accreta, with one study limited analysis to cases with histopathologically confirmed accreta,²⁵⁸ two used histopathologic diagnosis or clinical findings of adherent placenta or difficult manual removal,^{236, 256} one used ICD-9 codes,²⁵² and two did not describe diagnostic criteria.^{249, 254}

Table 19

Overall incidence of placenta accreta.

Using the designation of placenta accreta as defined by the authors, the incidence of placenta accreta increased with increasing number of cesarean deliveries (Table 20). The risk was not statistically significant until women had at least two prior cesarean deliveries. Women with one prior cesarean delivery had a rate of accreta of 0.3 to 0.6 percent. In comparison to women with no prior cesarean delivery, the odds ratio for accreta was 1.3 to 2.16, which was not statistically significant. The incidence of accreta rose with increasing prior cesarean deliveries from 1.4 percent in women with two or more prior cesarean deliveries to 6.74 percent for women with five or more prior cesarean deliveries. These results were statistically significant in all three studies. The odds ratio increased from 8.6 to 29.8.

Table 20

Incidence of placenta accreta by number of prior cesarean deliveries.

A statistically significant relationship between placenta previa and placenta accreta in women with prior cesarean delivery was noted in two studies.^{236, 258} As the number of prior cesarean deliveries rose, the presence of placenta previa increased the likelihood of placenta accreta from 3.3 to 4 percent in women undergoing their first cesarean delivery to 50 to 67 percent in women with four or more prior cesarean deliveries. Women with accreta had a statistically significant increased risk of hysterectomy (odds ratio 43 to 99.5). Additional maternal outcomes such as surgical injury, hemorrhage, transfusion, and death and neonatal outcomes were reported inconsistently such that meaningful analysis of an association between accreta and prior cesarean delivery was not possible (Table 21).

Table 21

Incidence of placenta accreta with placenta previa by number of prior cesarean deliveries.

One of the major limitations in analyzing studies regarding placental abnormalities was the lack of consistent definition among studies, especially for abruption, which may have resulted in misclassification. There was also the potential of surveillance bias as women with prior cesarean delivery may have had additional ultrasounds or observation at the time of delivery in anticipation of possible placental complications compared with women without known risk factors. Studies that used histopathologic diagnosis of accreta were therefore limited to hysterectomy patients and may have missed patients managed with conservative therapy. The majority of studies relied on retrospective data analysis and are therefore limited by the quality and consistency of the original data collection.

This review confirms prior reports of increasing incidence of accreta in women with previa depending on number of prior cesarean deliveries.²⁶² evaluated previa, accreta, and prior cesarean delivery in a 1985 paper which was not included in this report due to case collection prior to 1980. In the 286 women with previa, the incidence of accreta for women with zero to four prior cesarean deliveries was 5, 24, 47, 40, and 67 percent, respectively. Interestingly, the incidence of previa and accreta by prior cesarean delivery was similar to later studies except for women with one prior cesarean delivery. Miller reported an incidence by number of prior cesarean deliveries of 4 percent, 14 percent, 23 percent, 35 percent, and 50 percent,²⁵⁸ and Silver found 3 percent, 11 percent, 40 percent, 61 percent, and 67 percent, respectively.²³⁶ Although the individual patient numbers in each study were limited, the consistent findings suggest that women with previa and prior cesarean delivery are at increased risk of placenta accreta and thus more likely to require hysterectomy at the time of delivery.

The incidence of placenta previa or accreta in women with one or two prior cesarean deliveries was less than 1.5 percent. The highest risk group was women with three or more prior cesarean deliveries with a risk for previa of 3.3 to 4.2 percent and accreta of 4.7 to 6.7 percent. In the women with previa, the incidence of hysterectomy was 50–67 percent, and the OR for a hysterectomy with accreta was 43–99.5.

Further studies need to be performed to better evaluate additional risk factors for the development of placenta accreta and surgical management to minimize uterine scarring. Women desiring large families should be counseled about the risks of abnormal placentation with multiple cesarean delivery. As the number of cesarean delivery continues to rise, continued evaluation needs to be performed to optimize management of women with abnormal placentation and minimize maternal and neonatal morbidity and mortality.

Summary of abnormal placentation. The risk of abruption for women with any prior cesarean ranges from 0.10 to 0.15 percent. The risk does not appear to increase with prior cesarean or number of prior cesarean deliveries. Women with a prior cesarean delivery had a statistically significant increased risk of placenta previa compared with women with no prior cesarean at a rate of 1.2 percent (95 percent CI: 0.8 to 1.5 percent). The incidence increased with increasing number of prior cesarean deliveries. A prior cesarean is a significant risk factor for maternal morbidity in women with previa. Compared with previa patients without a prior cesarean delivery, women with one prior cesarean and previa had a statistically significant increased risk of blood transfusion (15 versus 32.2 percent), hysterectomy (0.7 to 4 percent versus 10 percent), and composite maternal morbidity (15 versus 23 to 30 percent). For women with three or more prior cesarean deliveries and previa, the risk of hysterectomy and composite maternal morbidity rose significantly (0.7 to 4 percent versus 50 to 67 percent and 15 versus 83 percent, respectively). The incidence of placenta accreta rose with an increasing number of prior cesarean deliveries. The results were statistically significant for women with two or more prior cesareans (odds ratio 8.6 to 29.8). Women with placenta previa are at increased risk for placenta accreta, and the risk increased with the increasing number of prior cesareans. Women with more than three prior cesareans and previa had a 50 to 67 percent incidence of accreta.

Maternal complications associated with multiple cesarean deliveries. As the number of women who attempt TOL decreases, the obvious consequence is an increase in the number of RCD. Conventional wisdom has suggested questioning women with a prior cesarean delivery about their plans for future childbearing as part of the discussion regarding mode of delivery due to the increased risk of multiple cesarean deliveries, but the actual risks remains unclear. Maternal morbidity resulting from multiple cesarean deliveries may consist of adhesions, hemorrhage/transfusion, surgical injury, postoperative infection, hysterectomy, abnormal placentation, and death.

The evidence regarding the outcome of multiple cesarean deliveries is limited and consists of 11 good or fair quality studies.^{81, 220, 235, 236, 249, 251, 252, 254, 257, 260, 263}

Hemorrhage. Three fair quality cohort studies evaluated the impact of multiple cesarean deliveries on maternal hemorrhage and/or blood transfusion rates (Table 22).^{249, 251, 263} Definitions of hemorrhage varied. Rouse et al used MFMU data to identify women who received a transfusion of packed red blood cells prior to hospital discharge.²⁵¹ Among women undergoing primary cesarean delivery, 3.2 percent (762/23486) received a blood transfusion. Of women with a prior cesarean delivery, the percentage of women with blood transfusions increased with increasing number of prior cesarean delivery from 1.8, 2.6, 4.3, 4.6, and 14.6 percent from one prior to five or more cesarean deliveries, respectively. The odds ratio for women with five or more cesarean deliveries was 7.6 (95 percent CI: 4.0 to 14.3). Nisenblat et al compared outcomes for women at a single institution in Israel undergoing a second versus three or more cesarean deliveries.²⁴⁹ Women were identified who experienced “excessive blood loss” of greater than 1000 mL or were transfused two or more units. Among women having their second cesarean delivery, 3.3 percent (16/491) met this definition compared with 7.9 percent (22/277) of those with two or more prior cesarean delivery (odds ratio 2.3; 95 percent CI: 1.1 to 4.5). Macones et al performed a secondary analysis of a multicenter, retrospective cohort study and examined incidence of blood transfusion.²⁶³ In women with two prior cesarean deliveries who did not attempt a TOL, 1.18 percent of 2,888 women received a transfusion (odds ratio 0.54; 95 percent CI: 0.23 to 1.27). These studies suggest that overall rates of hemorrhage/transfusion are less than 5 percent but the risk appears to increase with increasing numbers of cesarean delivery.

Table 22

Multiple cesarean deliveries effect on rates of hemorrhage.

Adhesions. Three good or fair quality studies discussed the presence of adhesions after multiple cesarean deliveries (Table 23).^{235, 249, 257} Uniform definitions of adhesions were not used. Nisenblat et al reported adhesions present in 25.6 percent (124/491) of women undergoing their second cesarean delivery versus 46.1 percent (124/277) for women with two or more prior cesarean deliveries (odds ratio 2.5; 95 percent CI: 1.8 to 3.4).²⁴⁹ Lynch et al studied outcomes at a single hospital in Ireland and found a similar rate of adhesions (48.8 percent) in women with two or more prior cesarean deliveries (122/250).²⁵⁷ Women with three or more prior cesarean deliveries at the University of Oulu, Finland were studied to identify outcomes in women with multiple cesarean deliveries.²³⁵ Records were reviewed for 64 women who underwent a total of 341 cesarean deliveries, 149 of which were their fourth or greater cesarean. These women were compared with a control group consisting of the next cesarean in the same situation (elective versus emergency). Intraperitoneal adhesions were noted in 18.2 percent of cases versus 2.7 percent of controls (odds ratio 8.1; 95 percent CI: 2.7 to 23.8). Overall, incidence of adhesions appears to increase with increasing numbers of cesareans.

Table 23

Multiple cesarean deliveries effect on rates of adhesions.

Surgical injury. The data regarding surgical injury and multiple cesarean deliveries is very limited and consists of two good quality studies (Table 24).^{236, 257} Both studies evaluated bladder injuries. Lynch et al found 1.6 percent of women with two or more prior cesareans had a bladder injury (4/250).²⁵⁷ Silver et al noted less than 0.3 percent of women with less than three prior cesareans experienced a bladder injury compared with 4.5 percent of women with five or more prior cesareans.²³⁶ This trend was statistically significant at p<0.001. Risk of bowel and ureteral injury with increasing number of cesareans was also statistically significant, although overall incidence was less than 1.2 percent. Bladder, bowel, and ureteral injury are uncommon occurrences and appear to increase with multiple cesareans.

Table 24

Multiple cesarean deliveries effect on rates of surgical injury.

Perioperative infection. The data regarding perioperative infection and multiple cesarean deliveries is limited and consists of four good or fair quality studies.^{81, 235, 236, 257} As indicated earlier in the report, there was no uniform definition of infection. Phelan et al reported an incidence of “febrile morbidity” of 19.2 percent (163/847) for women undergoing RCD, but the authors did not define febrile morbidity.⁸¹ Similarly, Juntunen et al. noted 14.1 percent of women with three or more prior cesareans had postoperative infections (odds ratio 0.9; 95 percent CI: 0.5 to 1.8), but the criteria for infection were not defined.²³⁵ Urinary tract infection (UTI) and upper respiratory tract infection (URI) were used by Lynch et al to describe postoperative infectious complications (Table 25).²⁵⁷ Silver et al defined postpartum endometritis clinically on the absence of findings consistent with an extrauterine source.²³⁶ There was a statistically significant increase in endometritis with multiple cesareans (p<0.001). Based on these studies, the risk of postoperative infection with multiple cesareans remains unclear.

Table 25

Multiple cesarean deliveries effect on rates of perioperative infection.

Wound complications. Two good quality studies report incidence of wound complications with multiple cesarean deliveries.^{236, 257} Silver et al reviewed wound infection and wound dehiscence and found no statistically significant change with multiple cesareans (p=0.09 and 0.18, respectively).²³⁶ Similarly, Lynch et al. found no correlation between number of cesareans and wound problems (Table 26).²⁵⁷

Table 26

Multiple cesarean deliveries effect on rates of wound complications.

Hysterectomy. Seven good or fair quality studies evaluated the rate of hysterectomy with multiple cesarean deliveries (Table 27).^{81, 236, 249, 252, 254, 257, 260} Women requiring hysterectomy due to abnormal placentation were discussed previously in this report (see Abnormal Placentation section). There are three population-based, matched case-control studies with women requiring peripartum hysterectomy chosen as cases.^{252, 254, 260} Bodelon et al used the Washington State birth certificate registry. Women undergoing their first cesarean delivery were more likely to require hysterectomy than were women delivering vaginally (odds ratio 4.6; 95 percent CI: 3.5 to 6.0). Women with one or more prior cesarean were significantly more likely to require hysterectomy (odds ratio 7.9; 95 percent CI: 5.8 to 10.7).²⁵² Knight et al used the United Kingdom Obstetric Surveillance System and similarly noted an increased risk for hysterectomy with primary cesarean (odds ratio 7.13; 95 percent CI: 3.71 to 13.7). The risk of peripartum hysterectomy for women with two or more prior cesareans was significantly higher (odds ratio 18.6; 95 percent CI: 7.67 to 45.4) than for women with one prior cesarean delivery (odds ratio 2.14; 95 percent CI: 1.37 to 3.33).²⁵⁴ Zelop et al used obstetric records at Brigham and Women’s Hospital to perform a case series of emergency peripartum hysterectomies between 1983 and 1991. Hysterectomy rates for women undergoing a primary cesarean were 0.062 percent, and increased with one prior cesarean delivery to 0.735 percent. Women with one or more prior cesarean had a hysterectomy rate of 1.08 percent, these rates were statistically significant.²⁶⁰ Nisenblat et al compared women undergoing a second cesarean versus women with two or more prior cesareans. The rate of hysterectomy increased from 0.2 percent (1/491) to 1.1 percent (3/277) in the multiple cesarean group, but the result was not statistically significant.²⁴⁹ Lynch et al found a similar rate of hysterectomy in women with four or more prior cesareans of 1.1 percent (2/170).²⁵⁷ Silver et al used MFMU data and noted increasing incidence of hysterectomy with increasing number of cesareans from 0.65, 0.42, 0.90, 2.41, 3.49, and 8.99 percent with zero to five or more prior cesareans, respectively. Women with five or more prior cesareans were 15 times more likely to require hysterectomy (odds ratio 15.2; 95 percent CI: 6.9 to 33.5), these results were statistically significant.²³⁶ These studies strongly support a correlation between multiple cesareans and hysterectomy. The odds ratio for hysterectomy with one prior cesarean is 0.7 to 2.14, with one or more is 1.4 to 7.9, and two or more is 3.8 to 18.6.

Table 27

Multiple cesarean deliveries effect on rates of hysterectomy.

Summary of maternal complications associated with multiple cesarean deliveries. Thirty-three percent of births in the U.S. as of 2007 were accomplished via cesarean delivery.¹ Many of these women will have additional children in the future and will be faced with the decision regarding mode of delivery. If she has a RCD, she will most likely have a cesarean for the remainder of her pregnancies. If she has a VBAC, she will likely have additional VBAC and avoid multiple cesareans. One of the fundamental gaps in the literature is intention for future pregnancies. This report supports prior evidence that although maternal morbidity increases with cesarean and neonatal morbidity increases with TOL, the overall incidence is extremely uncommon. The overwhelming majority of second pregnancies will result in a healthy mom and baby regardless of delivery method. This changes for women with multiple cesareans. As the number of prior cesareans increases, the maternal morbidity increases, especially for women with more than three prior cesareans. These women are at statistically significant increased risk of previa, accreta, and hysterectomy. The highest risk group is women with previa and prior cesarean, and the risks increase with increasing number of prior cesarean. Women with three or more prior cesareans and previa had a statistically significant increased risk of accreta (3.3 to 4 percent versus 50 to 67 percent), hysterectomy (0.7 to 4 percent versus 50 to 67 percent), and composite maternal morbidity (15 versus 83 percent) compared with women with previa and no cesarean. The only identified prevention of previa is avoiding uterine instrumentation. The overall incidence of previa is uncommon. The incidence of previa in women with any prior cesarean was 1.2 percent and for women with three or more prior cesareans it was 2.8 percent. There is no identified method for determining which women will develop previa in a subsequent pregnancy. All pregnant women are at risk for previa, women with previa are at increased risk of maternal morbidity, the incidence of previa and risk of morbidity increases with increasing number of prior cesareans, and there is no ability to predict which women will develop these complications. This has substantial implications for VBAC counseling relating to the risks of major morbidity associated with multiple cesareans in future pregnancies, especially for women desiring large families. Unfortunately, women are often unable to predict how many children they will have. Per the CDC (http://www.cdc.gov/reproductivehealth/UnintendedPregnancy/index.htm), the unintended pregnancy rate in the U.S. in 2001 was approximately 50 percent. Therefore, it is not unlikely that women will be facing additional pregnancies following cesarean, even if they were not planning to have more children. The inability to determine which women have completed childbearing and this report’s conclusion that maternal morbidity increases with multiple cesareans supports the ACOG 2004 practice bulletin recommendation that most women with a prior cesarean should be counseled about VBAC and offered a TOL.

Direction of scar. There are three recognized types of uterine incision; the most common approach is the low transverse incision through the lower uterine segment. Some surgeons prefer the low vertical incision, although concern remains that it may enter the muscular portion of the uterus. The classical incision, or high vertical incision, has been strongly associated with increased risk of uterine rupture and is a recognized contraindication to labor. Modifications of low transverse incisions—which require entering the muscular portion of the uterus, known as “T” or “J” incisions—are considered classical incisions for classification purposes. In addition, many women are unaware of what type of incision they had with a prior cesarean delivery, especially if the delivery was performed in countries other than the U.S. where operative reports may not be available. These incisions are classified as unknown. It remains unclear the effect of low vertical or unknown incisions on the risk of uterine rupture.

Impact of direction of scar. The published literature regarding impact of the direction of scar is limited and consists of one good and seven fair quality studies (Table 28).^{80, 84, 175, 203, 233, 264–266} These studies focus on impact of direction of scar and subsequent uterine rupture. As previously defined in this report, complete uterine rupture is a separation through the entire thickness of the wall including visceral serosa (with or without extrusion of part of all of fetal-placental unit). An incomplete uterine rupture is defined as a separation that was not completely through all layers of the uterine wall (e.g., serosa intact). Only one author used these definitions.⁸⁰ All other studies will therefore be listed as “uterine defect” which will encompass both complete and incomplete uterine rupture.

Table 28

Impact of direction of scar.

Uterine rupture. Landon et al 2006 used the MFMU database to study the risk of uterine rupture based on direction of scar for women attempting a TOL.⁸⁰ For women with a prior LTCD in that study, the overall rupture rate was 105/14,483 deliveries (0.7 percent). The rate with a prior LVCD was 2.0 percent (2/102 deliveries). There were also two uterine ruptures in 102 women with prior classical, J, or T incisions who refused cesarean delivery or presented in advanced labor, resulting in a rupture rate of 1.9 percent.

Uterine defect, prior low vertical incision - trial of labor. Three fair quality cohort studies reported on prior LVCS and incidence of uterine defects in women undergoing TOL.^{84, 175, 203} Paul et al examined 16,200 deliveries, of which 751 women had a prior cesarean delivery and attempted a TOL.¹⁷⁵ Fifty women had a prior LVCD, and per the authors, there were no serious complications. Martin et al studied 717 patients with prior cesarean who were separated into groups based on intended mode of delivery.²⁰³ For women who elected a TOL, six had a prior LVCD with no cases of uterine defects. Ninety-five women with a prior LTCD had VBAC deliveries with one uterine defect. Of the 61 women who required abdominal delivery after TOL, six had a prior LVCD with no cases of uterine defects, and 55 women had prior LTCD with four defects. The authors concluded that the evidence did not support the theory that low-vertical incisions are more likely to rupture than LTCD. Stovall et al performed a prospective cohort study of 272 women with a prior cesarean attempting a TOL at the University of Tennessee College of Medicine.⁸⁴ One hundred and fifty-nine out of 208 women with a prior LTCD underwent VBAC deliveries with one uterine defect. Fifty-seven out of 64 women with a prior LVCD underwent VBAC deliveries with no defects. These studies suggest that TOL with a prior LVCD does not have an increased rate of uterine dehiscence compared with prior LTCD, but the total number of TOL attempts with prior LVCD reported in these three studies is 126. This suggests that the sample size may have not been large enough to capture the true risk of uterine defect with a prior LVCD, but it is reassuring that there were no cases of uterine defects in this series.

Uterine defect, prior low vertical incision - elective repeat cesarean delivery. Martin et al studied 717 patients with a prior cesarean delivery separated into groups based on intended mode of delivery.²⁰³ Of the participants, 547 chose ERCD, 483 with a prior LTCD and 64 with a prior LVCD. There were six uterine defects in the LTCD group and no cases in the LVCD group.

Uterine defect, prior low vertical incision - unknown intended mode of delivery. Three fair quality cohort studies evaluated a prior LVCD and incidence of uterine defect but did not discuss intended mode of delivery.^{233, 265, 266} As the authors did not report which patients attempted TOL, it is unclear whether patients with scar disruption presented with asymptomatic defects at the time of ERCD or after attempted TOL. Tahilramaney et al reported defects in 2.8 percent of 374 patients with prior LTCD.²⁶⁶ One hundred and thirty-four women delivered vaginally with uterine defects in three versus 211 delivering via cesarean with seven cases of defect. For the 11 patients with a prior LVCD, all were delivered via cesarean with one defect (9 percent). For the 21 prior classical or vertical incisions of unknown type, there were five vaginal deliveries without complications and two cases of uterine defects with cesarean (14 percent). The authors concluded that type of uterine incision was not statistically significant in regards to uterine dehiscence. Leung et al reviewed cases of uterine defects at one institution over a 10 year period.²³³ There were 16,467 women who had a prior cesarean with 107 cases of uterine defects, 99 with complete records. Per patient history, 90 percent of scar types were unknown, but, when possible, scars were classified at the time of laparotomy. Eleven were recorded as classical/vertical, 64 were transverse, and 24 remained unknown. A database at Emory University was used by Lin et al to identify patients with a prior cesarean who delivered at greater than 28 weeks and to study the impact of the direction of scar and uterine defects.²⁶⁵ Of 3,533 patients, 145 had a prior classical scar, 1,931 had a LTCD, 39 a LVCD, and 1,312 had an unknown scar. There were 106 patients excluded for unidentified scar. There were no cases of uterine defects with prior LVCD or classical incisions.

Uterine defect, unknown uterine scar. Unknown uterine scar remains a diagnostic challenge. One good and three fair quality studies evaluated outcomes for women with unknown incisions.^{221, 264, 266} Grubb et al performed a RCT of 197 women in latent labor with unknown uterine scars comparing nonintervention to active management.²⁶⁴ In the intervention group, there were five cases of uterine defects. There were no cases in the nonintervention group (0 versus 5 percent, p=0.03). Per the author’s definition, there was one case of rupture and four uterine dehiscences. The uterine rupture was through a vertical scar (later called a T incision). The four cases of uterine dehiscence were with LVCD, and three were noted on routine exploration of the uterine cavity following VBAC. Tahilramaney et al reviewed 451 patients with unknown incisions, 93 delivered vaginally with no complications, and 319 delivered via cesarean delivery, of which 11 (2.6 percent) experienced uterine defects.²⁶⁶ Lin et al reviewed 1,312 patients with unknown scar. In comparison to patients with a known prior LTCD, there was no increased rate of uterine defects in patients with an unknown scar (0.6 and 0.5 percent, respectively).²⁶⁵ Landon et al found that for women with an unknown scar, there were 15 ruptures in 3,206 deliveries (0.5 percent).²²¹ These studies suggest that women with an unknown scar are not at significantly increased risk of uterine dehiscence or rupture with TOL.

Direction of scar summary. Because the scope of this report started after the NIH conference in 1980, data regarding the risk of uterine defect for classical incision is largely absent and what is there is likely biased as providers in general will not allow a trial of labor among women with prior classical incisions. Studies prior to 1980 suggest that women with prior classical are at substantially increased risk for uterine rupture and should not undergo labor. The evidence regarding prior LVCD is very limited. Of six studies on 336 women with LVCD, there are two reported cases of uterine rupture and one uterine defect. These limited data suggests that women with a prior LVCD are not at a significantly increased risk of uterine dehiscence or rupture compared with women with a prior LTCD. Women with an unknown scar are not at a significantly increased risk of uterine dehiscence or rupture with TOL compared with women with prior LTCD.

Obesity. Because obesity is an increasingly important health problem in the U.S., the impact of BMI and/or weight on VBAC rate was investigated. RCT, cohort, or case series studies that reported weight or BMI for TOL or ERCD groups for VBAC rate, or maternal or infant outcomes were included. Of the 119 full text articles retrieved and assessed for inclusion, seven good or fair quality cohort studies were reviewed.^{78, 110, 126, 135, 240, 267, 268} A number of studies provide context for the data related to maternal BMI from the MFMU cohort,^{80, 221}, one highlighting health outcomes by BMI and TOL versus ERCD groups will be discussed here.⁷⁸ Three studies of poor quality were excluded from analysis.^269–271 Most studies stratified BMI in predefined categories,^{78, 110, 126, 267} while others used weight.^{135, 240, 268} The most commonly used definition included four BMI (kg/m2) categories: normal, less than 25; overweight, 25 to 29.9; obese, 30 to 39.9; and morbidly obese greater than 40. BMI cutoffs did vary within categories. One looked only at women eligible for a TOL who weighed more than 300 pounds.²⁴⁰

Overall VBAC rates compared by BMI groups will be discussed (Table 29) along with a brief overview of TOL maternal and infant outcomes by BMI status (Table 30). The largest of the U.S. cohorts will be emphasized as it provides a comprehensive look at how morbid obesity may affect maternal and infant morbidity, TOL compared with ERCD.⁷⁸

Table 29

Vaginal birth after cesarean rate in studies with data by body mass index.

Table 30

Summary of findings on maternal and infant health outcomes for trial of labor and body mass index.

Vaginal birth after cesarean rate and body mass index. Overall VBAC rates ranged from 66 to 79 percent in the included obesity studies.^{78, 110, 126, 135, 267, 268} When looking at VBAC rate by BMI, most studies found a greater percentage of normal weight women achieved VBAC when compared with overweight or obese women.^{78, 110, 126, 135, 267, 268} The three studies reporting VBAC based upon BMI of greater than or equal to 40 showed varying VBAC rates: 52.1 percent,²⁶⁷ 61 percent,⁷⁸ and 70 percent.¹²⁶ The two smallest U.S. studies that used weight rather than BMI showed women over 300 pounds had a 13 percent VBAC rate.^{240, 268} In a U.S. retrospective chart review, linear regression revealed that as BMI increases, the VBAC rate decreases (r=−0.182, P=0.001).¹³⁵ The MFMU study shows a relationship between increasing BMI and decreasing VBAC rate where normal BMI women had an 85 percent VBAC rate while those categorized as morbidly obese had a 61 percent VBAC rate (p<0.001; Table 29).⁷⁸

The MFMU⁷⁸ enrolled over 28,000 eligible women aiming to better understand BMI and the risks of uterine rupture and infant and maternal morbidity associated with a TOL as compared with ERCD. BMI was calculated at delivery using weight in kilograms divided by the square of the height in meters (kg/m²). The subanalysis includes term (greater than or equal to 37 weeks GA), singleton pregnancy with a prior cesarean delivery and a BMI greater than 18.5kg/m² at delivery. The description of the cohort reports a significantly higher BMI in the ERCD group compared with the TOL group (BMI: 32.1+6.6 versus 33.6+7, p<0.001).⁷⁸ Health outcomes for those with morbid obesity (BMI greater than 20) were compared for TOL and ERCD.

Outcome data from the MFMU TOL versus ERCD groups provide insight into how morbid obesity may contribute to adverse pregnancy outcomes for both mother and infant. Compared with the ERCD group, the TOL group had a greater likelihood of hospital stay (odds ratio 1.2; 95 percent CI: 1.1 to 1.4), endometritis (odds ratio 2.4; 95 percent CI: 1.7 to 3.5), dehiscence (odds ratio 2.4; 95 percent CI: 1.0 to 5.4), rupture/dehiscence (odds ratio 5.6; 95 percent CI: 2.7 to 11.7), composite morbidity (odds ratio 1.2; 95 percent CI: 1.1 to 1.4) composite morbidity (excluding stay, odds ratio 1.8; 95 percent CI: 1.5 to 2.6).⁷⁸ TOL did not have higher risk of transfusion, maternal surgical injury, hysterectomy, wound complications, or thromboembolic disease. Compared with the ERCD group, the TOL infants were at greater risk for 5 minute Apgar scores less than seven (odds ratio 3.1; 95 percent CI: 2.1 to 4.6) and injury (odds ratio 5.1; 95 percent CI: 1.9 to 13.8), but not 5 minute Apgar score less than three, sepsis, NICU admission, or stillbirth/abortion/neonatal death.⁷⁸

Health outcomes by body mass index. Table 30 provides a catalog of statistical findings for those included obesity studies that provide information on maternal and infant health outcomes by BMI or weight status. Data are summarized for between BMI group differences in rates of adverse outcomes as reported in full text papers, with statistical significance also provided.

Maternal health outcomes. Three cohort studies provide information on women experiencing uterine rupture or dehiscence between the BMI groups.^{78, 135, 267} One German cohort study (N=8580) found no difference by BMI group for uterine rupture separation,²⁶⁷ while the largest MFMU study found no differences when uterine rupture and dehiscence were analyzed alone, but when combined, a significantly higher rate in the largest of four BMI groups was found (groups/percent 1) 0.9, 2) 1.5, 3) 1.4, 4) 2.1, respectively, p=0.03).⁷⁸ Another study reported uterine rupture rates were higher in the overweight group (3.6 percent) than in the underweight (0.6 percent), normal (1.8 percent), and obese (0 percent) groups, (p=0.041); however, when controlling for number of layers of closure, this no longer held true (odds ratio 5.08; 95 percent CI: 0.53 to 48.79, p=0.159).¹³⁵

More overweight and obese women compared with normal weight women had trouble with wound healing complications in the included studies. Compared with lower BMI groups, women with greater BMI had statistically significantly more wound infection and fever,²⁶⁷ wound infection or endometritis,²⁶⁸ or wound complications and endometritis.⁷⁸ Alternatively, one U.S. cohort (N=725) did not find a significant difference between BMI groups for infection.¹²⁶ While no differences in surgical injury rates were reported in the MFMU,⁷⁸ those women with lower BMI had significantly more third and fourth degree lacerations than women with higher BMI in another study (p<0.01). Studies reporting hysterectomy data did not find any differences between BMI groups.^{78, 267} In addition, hemorrhage,²⁶⁷ blood loss,²⁶⁸ and transfusion⁷⁸ showed no BMI group differences. The two studies looking at maternal hospital stay found statistically significantly higher rates of hospital stay with higher BMI.^{78, 268} The MFMU sub analysis showed maternal stay of 4 or more days in the hospital at 30.3 percent in the morbidly obese BMI category, versus 9.4 percent, 12 percent, and 18.9 percent, in the other weight categories respectively, p<0.001.⁷⁸ Finally, there was no significant difference in the adverse event thromboembolism in BMI groups.⁷⁸

Infant health outcomes. Two studies reporting on infant death did not find a statistically significant difference between BMI groups.^{78, 267} The MFMU secondary analysis did note that stillbirth/abortion and neonatal death were the highest in the morbid obesity group (BMI greater than 40 was 0.5 percent versus 0.2, 0.3, and 0.3, in the other weight categories, respectively, p=0.14).⁷⁸ When comparing with lower BMI categories, neonatal resuscitation with intubation²⁶⁷ and NICU admission⁷⁸ were highest in the morbidly obese groups as well (greater than 40 BMI) (p=0.03 and p<0.001, respectively).

GA and birth weight were common outcomes evaluated in all studies with BMI group data; however, results were mixed. Two studies showed no effects for GA,^{135, 268} while two studies indicate that younger GA is associated with lower BMI.^{126, 267} The largest MFMU study found older GA in the overweight and obese BMI groups, when compared with the underweight or normal BMI groups (p<0.001).⁷⁸ Similarly, a U.S. university clinic cohort found older GA in higher BMI patients, compared with lower BMI patients.¹²⁶ A similar picture emerges with infant weight and maternal BMI, where three large U.S. studies showed infant weights greater than 4,000 grams were more likely in the largest BMI categories,^{126, 135, 267} and two where infant weight was higher in high BMI groups.^{78, 126} In contrast, one study reported no infant weight differences in BMI groups.²⁶⁸

Apgar scores less than seven taken at 1 minute following delivery showed no differences by BMI,^{126, 267} while Apgar less than seven taken at 5 minutes were more likely in the high BMI group compared with lower BMI group.²⁶⁷ The MFMU analysis showed 5 minute Apgar scores less than or equal to three were not significantly different between BMI groups.⁷⁸ In addition, infant injury and sepsis were similar for all maternal BMI groups.⁷⁸

Summary of obesity. Assessing the risks and benefits of VBAC using BMI is a complex exercise. Data show that increasing BMI is linked to decreased VBAC, with morbidly obese women and their infants being at the highest risk for adverse outcomes. Compared with normal weight mothers, data suggest that obese and morbidly obese women are more likely to suffer rupture and/or dehiscence, wound infection, and/or increased hospital stay, while their infants may experience more injury and greater weights, specifically greater than 4,000 grams. BMI categories use different cut offs and patient inclusion and exclusion criteria vary. Comparison across studies for VBAC rate and important health outcomes is further complicated by lack of consensus on definitions and priorities. Future research in community practice settings with reproducible and valid outcome measures could provide more insight in this field.

Effects of management of trial of labor using a protocol on maternal outcomes. In a small cohort study examining the impact of using a strict protocol for managing a trial of labor in 841 women with prior cesarean delivery compared with 467 women undergoing ERCD maternal harms were evaluated.⁹¹ The rate of major complications (uterine rupture, hysterectomy, relaparotomy, operative injury, or greater than two units of blood transfused) was not statistically significantly different between the TOL and ERCD groups (1.8 versus 1.3 percent, p=0.50).

Adhesions

One concerning complication from multiple cesarean deliveries is increased complications from adhesive disease. This may result in a more difficult RCD, increased postoperative complications, or increased complications with future gynecological surgeries. Studies looking at adhesions are limited. One study described increased adhesions with increased number of cesarean deliveries at the time of cesarean.²³⁵ Another study found increased perioperative complications with vaginal hysterectomy to be associated with women who underwent one or more cesarean in the past (frequency 18.31 versus 3.58 percent, p<0.0001). Specifically, in women with a history of cesarean, there was a 5.63 percent rate of bladder injuries compared with 0.89 percent in women without cesarean delivery (p= 0.01). A history of cesarean was also associated with increased need for adhesiolysis, intestinal injuries, and longer operating time, but none of these factors were significant. This study did not delineate one versus multiple cesareans, however, and did not evaluate patients with a history of VBAC.²⁷²

A similar study evaluating factors relating to complications during hysterectomy found a history of cesarean delivery to be significantly associated with an increase in complications. Again, however, this study did not evaluate multiple cesareans or VBAC separately.²⁷³ Another study evaluated the benefit of closing the peritoneum at the time of initial cesarean to decrease the incidence of adhesions in future cesareans. This study found a statistically significant increase in the number of patients with severe adhesive disease in those without peritoneal closure compared with those who had closure at the time of RCD. (6 versus 42 percent, p=0.003) This study did not report the rate of severe adhesions in those patients undergoing multiple RCDs; however it does provide evidence that even one cesarean can result in significant adhesions. The presence of adhesions can result in complications with subsequent cesareans.²⁷⁴ This finding was supported by a Canadian study evaluating the presence of adhesions in subsequent cesarean. This study found a dose dependent relationship with the prevalence of adhesions with the number of cesareans, from 0 percent at primary cesarean delivery to 47.9 percent at the fourth cesarean.²⁷⁵ In addition, this study found a statistical increase in both delivery time as well as total operative time for deliveries complicated by adhesions compared with those without adhesions. Operative and delivery times did not increase however with increasing number of cesareans.²⁷⁵

One study evaluating the incidence of postoperative small bowel obstruction, thought to be associated with increased adhesive disease, found a smaller rate of obstruction in the cesarean delivery group than among those with hysterectomy, adenexal surgery, or myomectomy (0.05 percent versus 1.63, 0.87, and 0.39 percent; p<0.001).²⁷⁶

Reproductive Health

Several studies have attempted to define the impact of cesarean delivery on overall reproductive health. One of the important factors emphasized by clinicians is the ultimate family plan of patients, due to the increased risk perceived with multiple cesareans. However, there is also a concern for impaired fertility due to surgery on the uterus. However, this is not well studied with only two studies evaluating impaired fertility. One study of fair quality looking at future pregnancies after vaginal delivery, operative vaginal delivery, and cesarean found there to be a statistically significant difference in the ability to conceive in subjects undergoing a cesarean compared with those who underwent an instrumented vaginal delivery (odds ratio 0.33; 95 percent CI: 0.12 to 0.98). This study however, did not distinguish between VBAC and primary deliveries in the cohort.²⁷⁷ The second study of fair quality found a history of cesarean associated with an increased odds of taking greater than one year to conceive after adjusting for confounders including maternal and paternal age, demographics, and BMI (odds ratio 1.53; 95 percent CI: 1.09 to 2.14). Though this study did not evaluate VBAC or multiple cesareans separately, there was an increased risk of delayed fertility with increasing parity, suggesting a continued effect of at least one cesarean (odds ratio 2.97; 95 percent CI: 1.72 to 5.10).²⁷⁸

One case-control study found an increased odds ratio of multiple cesarean deliveries (greater than or equal to two priors) compared with no pelvic surgery in women with early menopause (odds ratio 2.69; 95 percent CI: 1.16 to 6.22) This study however, did not evaluate VBAC and this association was not evaluated in a multivariate model.²⁷⁹ No studies evaluated TOL and/or RCD with respect to pelvic pain, risk of ectopic pregnancy, and general health risks, such as diabetes or high blood pressure.

Perinatal Mortality

The definitions accepted by the National Center for Vital Statistics⁶³ were used to review and describe the data relating to perinatal mortality and the subsets of fetal and neonatal mortality in women with a prior cesarean delivery. The definition of perinatal mortality (perinatal II) included infants less than 28 days of age and fetal deaths of 20 weeks or more gestation. To study the frequency of stillbirth (antepartum and intrapartum) we used both the intermediate (20 to 27 weeks gestation) and late (28 weeks or greater gestation) fetal definitions of fetal death in an attempt to capture the most studies and allow comparisons to national statistics for the general population. Studies that reported fetal loss less than 20 weeks gestation or less than 500 grams were not included in the review. Neonatal (infant) mortality was defined as death in the first 28 days of life.⁶³. To reduce the effects of prematurity on the neonatal mortality rate, we limited our analyses of neonatal mortality to term infants. Two studies,^{280, 281} which focused specifically on the risk of a stillbirth in a subsequent pregnancy after prior cesarean delivery (irrespective of mode of delivery in the next pregnancy), were included and are discussed in the section of long-term outcomes and the impact of the mode of delivery on subsequent pregnancies. Eight cohort studies of good or fair quality and reporting data on mortality using at least one of these definitions of mortality are included.^{79, 93, 103, 204, 228, 282–284}

Perinatal morality rate. The U.S. perinatal mortality rate (PMR) for infants 28 weeks gestation to less than 7 days of life was reported to be 0.66 percent for the year 2005, but notably does not exclude infants with congenital anomalies.⁶³ Five good or fair quality cohort studies, involving 76,899 infants reported perinatal mortality associated with TOL and ERCD (Table 31).^{103, 204, 228, 282, 284} The definition of perinatal mortality among the five studies included fetal and neonatal deaths up to 28 days of life. Perinatal mortality was also used to categorize studies if it was unclear if the death occurred during labor or after delivery.²⁸⁴ All five studies focused exclusively on women delivering at term. Three of the studies occurred in tertiary or university settings^{204, 228, 284} and two utilized population databases.^{103, 282}

Table 31

Perinatal mortality rate (20 weeks or greater gestation to 7 days of life) among any gestational age studies.

There were 72 perinatal deaths/41,213 births in women having a TOL. The combined PMR for women undergoing a TOL was 0.13 percent (95 percent CI: 0.06 to 0.3 percent), translating to 1.3 per 1,000 (95 percent CI: 0.6 to 3 per 1,000). There were 46 perinatal deaths/35,686 births for women undergoing an ERCD. The combined PMR for ERCD was 0.05 percent (95 percent CI: 0.007 to 0.38 percent) this translates to 0.5 per 1,000 (95 percent CI: 0.07 to 3.8 per 1,000 (Figure 27)). The risk of perinatal mortality was significantly higher for TOL as compared with ERCD (RR 1.82; 95 percent CI: 1.24 to 2.67; p=0.041). Using 0.05 percent as the baseline risk for ERCD, the calculated risk difference was 0.41 percent (95 percent CI: 0.012 to 0.08 percent) which is equivalent to .41 more deaths among women who attempt TOL. One study²⁰⁴ examined the influence of labor and underlying maternal medical complications (indications) upon perinatal mortality. The PMR was higher among women with underlying medical conditions (indications) with and without labor compared with women without indications. Interestingly, the impact of labor upon perinatal mortality appeared to differ based upon indication status with PMR being higher among the labor group for women without indications (0.22 percent labored versus 0.12 percent no labor) and lower for the labor group for women without indications (0.19 percent labor versus 0.34 percent no labor).²⁰⁴

Figure 27

Perinatal mortality rate for trial of labor versus elective repeat cesarean delivery among all studies. *95% CIs are exact confidence intervals.

Fetal mortality rate. The fetal mortality rate (FMR; 20 weeks gestation to birth) in the U.S. for the year 2005 was 0.622 percent of live births, leveling off in 2004 (0.620 percent of live births) after two decades of decline from a rate of 0.783 percent of live births.⁶³ To understand the relationship of a prior cesarean delivery on fetal mortality (antepartum and intrapartum death) in women who undergo a TOL versus an ERCD, two good quality studies of women at term^{103, 204} met the inclusion criteria and were reviewed. One study used a retrospective cohort from an administrative database,¹⁰³ while the other used a prospective cohort design.²⁰⁴ The studies also differed in that one study¹⁰³ excluded fetal death prior to the onset of labor. Neither study limited their analyses by number or direction of uterine scar. Overall, the FMR was low in women attempting a TOL in both studies. When comparing only intrapartum stillbirth, the rates of intrapartum fetal demise (IUFD) ranged from 0.01 to 0.04 percent of live births in women attempting a TOL to 0 to 0.004 percent for women having an ERCD. Additionally, Spong (2007) measured antepartum stillbirth and found a rate of 0.21 percent in women undergoing a TOL versus 0.1 percent in women having an ERCD (Table 32).²⁰⁴

Table 32

Fetal mortality rate (20 weeks or greater gestation and before birth) among term studies.

Neonatal morality rate. Six good or fair quality cohort studies reported on the neonatal mortality rate (NMR) in women undergoing a TOL versus an ERCD (Table 33).^{79, 93, 103, 204, 228, 283} There was a wide range of hospital settings among the six studies, with two studies representative of academic medical centers,^{204, 228} two studies representative of population databases,^{92, 103} and two studies representative of a diversity of hospital types.^{79, 283} Overall, the neonatal mortality rate for TOL was low with a total of 51 neonatal deaths in a total of 44,485 subjects, for a combined NMR of 0.11 percent (95 percent CI: 0.06 to 0.2 percent). A total of 40 neonatal deaths occurred in 63,843 women who had either an IRCD or ERCD for a combined NMR of 0.6 percent (95 percent CI: 0.02 to 0.15 percent) in the cesarean delivery group (Figure 28). The risk of neonatal mortality was significantly higher for TOL compared with ERCD with a calculated risk difference of 0.058 percent (95 percent CI 0.019 to 0.117 percent), which is equivalent to .58 additional perinatal deaths per 1,000 for TOL.

Table 33

Neonatal mortality rate (death occurring in the first 28 days of life) among term studies.

Figure 28

Neonatal mortality rate for trial of labor versus elective repeat cesarean delivery among term studies. *95% confidence intervals are exact

Two studies provided insight into classifications of patients who appeared to have higher NMRs when compared with subgroups.^{93, 204} In women who were classified as having high-risk conditions,⁹³ the NMR in the TOL group was 0.38 percent compared with women without high-risk maternal conditions undergoing a TOL (0.13 percent) and women undergoing ERCD (high‐risk women: 0.1 percent NMR; no high-risk condition: 0.05 percent NMR). In another study representing the MFMU cohort, Spong et al classified cesarean delivery by IRCD (labor versus no labor) and ERCD (labor versus no labor). In this individual study, the NMR was highest in the IRCD (no labor) at 0.2 percent.²⁰⁴ Overall, there was no difference in the TOL NMR (0.08 percent) versus overall cesarean delivery group NMR (0.08 percent).

While overall rates of perinatal mortality are lower in the reviewed studies in comparison with U.S. data for the general population, a noticeable pattern is the association between country of origin and perinatal death. The three studies conducted in the U.S. ^{79, 93, 204} reported higher perinatal, fetal, and neonatal mortality, particularly among TOL patients, compared with studies conducted outside of the U.S. The U.S. does have a higher infant mortality rate compared with either the United Kingdom or Canada (0.626 percent, 0.504 percent and 0.485 percent, respectively).²⁸⁵ The studies do not consistently provide details about demographic, societal, or health systems issues to explore the potential contributors.

Summary and strength of the evidence on perinatal death. Overall, the strength of evidence on perinatal mortality was low to moderate. The perinatal, fetal, and neonatal mortality rates reported were low, especially when compared with U.S. perinatal statistics from the CDC. However, CDC perinatal mortality data do not exclude congenital anomalies. While overall, perinatal, fetal, and neonatal mortality rates are low in women with a history of prior cesarean delivery, the death rates are significantly higher in women who attempt a TOL versus ERCD. Women with high-risk conditions and IRCD appear to have higher rates of neonatal mortality.

Infant Morbidity

Because of the association between infant outcomes and prematurity, studies of neonatal morbidity were limited to term neonates and included 11 studies of good or fair quality.^{79, 80, 90, 93, 97, 103, 156, 204, 283, 284, 286}

Respiratory conditions. Respiratory disorders in newborns are common and account for the majority of admissions to the neonatal intensive care unit (NICU) in the immediate newborn period.²⁸⁷ While generally considered an intermediate outcome, the need and level of required respiratory support at birth is an outcome of clinical, parental, and economic interest. Respiratory morbidity can occur regardless of mode of delivery, making conclusions about the relationship to the method of labor and delivery unclear. Determining which respiratory indicators are most representative of morbidity as well as disagreement about the definitions creates further challenges.

While respiratory distress syndrome (RDS) is primarily a disease of prematurity, term neonates can experience respiratory issues at or immediately following birth. Comparisons of studies are challenged by: (a) lack of standardized or mutual agreement on definitions of respiratory conditions; (b) lack of clarity regarding the importance and clinical significance of measures; (c) differences in birth settings; and (d) experience and skill level of available providers and staff. It is also worthwhile to note that significant practice changes have occurred over the course of the past several decades (e.g., the revised approaches to suctioning of an infant with meconium-stained amniotic fluid). These changes make it difficult to draw meaningful conclusions about measurements such as the presence of a pediatrician at delivery or the frequency of intubation for meconium without specific information about the measurement, context, and terminal measure of intermediate variables.

Six term fair quality cohort studies^{79, 90, 93, 97, 284, 286} reported an array of respiratory symptoms and interventions in the neonates of women who underwent TOL after cesarean delivery versus ERCD. Several studies made direct comparisons challenging in that the authors grouped respiratory symptoms or disorders together.^{90, 93, 97} Studies that measured the frequency of transient tachypnea of the newborn (TTN), bag-and-mask ventilation, intubation for meconium and ventilation in infants born after a TOL versus ERCD are found in Table 34.

Table 34

Comparison of cohort studies reporting respiratory morbidity in infants among term studies.

Three studies^{79, 90, 286} compared the frequency of bag-and-mask ventilation in the infant when women underwent TOL versus ERCD (Figure 29). The summary estimate of rates for infants needing bag-and-mask ventilation for TOL was 5.4 percent (95 percent CI: 3.5 to 7.6 percent) while the rate for ERCD was 2.5 percent (95 percent CI: 1.6 to 3.6 percent). Infants in the TOL group were significantly more likely to receive bag-and-mask ventilation with a pooled RD for TOL versus ERCD of 2.5 percent (95 percent CI: 0.72 to 5.0 percent.). The I² statistic for heterogeneity was 42.9 percent (between study heterogeneity accounts for 42.9 percent of the total heterogeneity). The Q-statistic for heterogeneity was 3.5 (p=0.1736).

Figure 29

Need for bag-and-mask resuscitation among term studies.

Three studies reported rates of TTN (Figure 30).^{79, 90, 284} There was significant heterogeneity among the three studies with a Q-statistic of 6.05 p=0.0485 and I² of 67 percent (indicating the between-study heterogeneity accounts for 67 percent of the total heterogeneity). The pooled absolute risk for TTN in the TOL group was 3.6 percent (95 percent CI: 0.9 to 8.0 percent) and 4.2 percent (95 percent CI: 1.9 to 7.3 percent) for ERCD. There was no statistically significant difference between the risk in the two groups using a random effects model with a pooled RD of −0.83 percent (95 percent CI: −3.35 to 1.7 percent) for TOL versus ERCD.

Figure 30

Rates of transient tachypnea of the newborn for trial of labor versus elective repeat cesarean delivery among term studies.

Only two studies^{79, 90} measured respiratory care of the infant with meconium-stained amniotic fluid. While one study⁹⁰ measured intubation for meconium, Hook et al⁷⁹ measured meconium aspiration syndrome. Both studies found respiratory care due to meconium-stained amniotic fluid to be greater in infants undergoing a TOL versus ERCD (Table 34).

One study examined respiratory morbidity occurring in infants born by planned and actual route of delivery (e.g., ERCD with and without labor, VBAC, and TOL resulting in RCD).²⁸⁶ This study found infants born after a TOL resulting in CD required the most bag-and-mask ventilation and intubation, while infants born by ERCD ( with our without no labor) required the most oxygen therapy (blow-by oxygen, continuous positive airway pressure).²⁸⁶

Ultimately, there were very few studies that reported on respiratory outcomes for term infants born by VBAC or ERCD. While infants born after a TOL were more likely to require bag-and‐mask ventilation compared with infants born by ERCD, other respiratory outcomes revealed no differences between routes of delivery. The fact that there are so few studies that measure respiratory outcomes for term infants is an important issue as respiratory outcomes are very important to clinicians and patients.

Summary and strength of the evidence on infant respiratory morbidity. The strength of evidence on the respiratory morbidity of the infant for VBAC versus ERCD was low due to lack of precision in estimates and inconsistency in findings. Respiratory distress syndrome was not included in this review of the literature because RDS is primarily a disease of prematurity and neonatal outcomes were primarily focused on term neonates to reduce the confounding by prematurity. Studies were conflicting regarding whether VBAC or ERCD resulted in more TTN. Two studies found significantly more infants required intubation for meconium in infants undergoing TOL versus ERCD. is a general lack of consensus among studies regarding what types of respiratory indicators are most representative of health or morbidity.

Hypoxic-ischemic encephalopathy/asphyxia. Hypoxic ischemic encephalopathy (HIE), neonatal encephalopathy, asphyxia, perinatal asphyxia, and hypoxia are terms used in the literature to describe a potentially serious neonatal complications. These descriptors attempt to link a hypoxic event during birth to intermediate and/or long-term neonatal outcomes. Several challenges exist in trying to capture the frequency and severity of such an outcome including lack of agreement regarding definition, timing, and measurement. Proposed criteria to define an acute intrapartum hypoxic event as sufficient to cause a long-term outcome such as cerebral palsy have been advanced by ACOG and the International Cerebral Palsy Task Force.²⁸⁸ The presence of four essential criteria have been proposed in order to link an intrapartum hypoxic-ischemic insult causing a moderate to severe neonatal encephalopathy resulting in cerebral palsy: 1) evidence of metabolic acidosis in fetal umbilical cord arterial blood obtained at delivery (pH less than 7 and base deficit of 12 mmol/L or more), 2) early onset of severe or moderate neonatal encephalopathy in infants born at 34 or more weeks' gestation, 3) cerebral palsy of the spastic quadriplegic or dyskinetic type, and 4) exclusion of other identifiable etiologies, such as trauma, coagulation disorders, infectious conditions, or genetic disorders.²⁸⁸

Three fair quality cohort studies attempted to measure this phenomenon in some manner.^{80, 93, 284} Gregory et al, in a large, population-based study, measured the frequency of hypoxia (by ICD-9 codes) in high- and low-risk women who underwent TOL and ERCD at term.⁹³ High-risk antepartum conditions contributing to neonatal hypoxia in women with a prior cesarean delivery were antepartum bleeding (odds ratio 2.9; 95 percent CI: 1.3 to 6.6) and oligohydramnios (odds ratio 2.5; 95 percent CI: 1.1 to 5.7). Overall, there was little difference in the frequency of hypoxic events (defined by ICD 9 category codes 768.0,1,2,3,4,5,6,9) in infants born to mothers with no high-risk clinical condition (TOL: 0.89 percent, ERCD: 0.32 percent) versus infants whose mothers had any number of high-risk clinical conditions (TOL: 1.29 percent versus ERCD: 0.20 percent). Landon et al, as part of the MFMU cohort, measured but did not offer a definition of HIE.⁸⁰ This study did not find the frequency of HIE to be significantly different in term women with one prior cesarean delivery compared with those with multiple prior cesarean deliveries (0.1 percent versus 0, p=1.0) or when comparing TOL versus ERCD in women with multiple prior cesarean deliveries (0 versus 0). Richardson et al used a very different approach from either of the other two and measured and reported on umbilical cord pH.²⁸⁴ This study found a slight decrease in the mean cord pH of women undergoing a TOL versus women who had an ERCD with no labor (7.24±0.07 versus 7.27±0.05, p< 0.001). There was a statistically significant difference in base excess in the TOL versus ERCD group (−5.3mmol/L ±3.1 versus −2.9mmol/L±2.5, p< 0.001). There were no significant differences in infants with umbilical artery pH of less than 7.00 in the same groups (TOL: 0.5 percent versus ERCD without labor: 0.1 percent, p=NS).

Summary and strength of the evidence on infant respiratory morbidity The strength of evidence on the HIE of the infant for VBAC versus ERCD was low due to lack of consistency in measurement and few studies. While the studies consistently report higher risk for HIE for TOL compared with ERCD, it is not possible to know the true relationship due to the low strength of overall evidence.

Sepsis

Three fair quality cohort studies were found to address sepsis in the neonates of women who attempted a VBAC versus ERCD.^{79, 90, 97} Fisler et al found that neonates in a university setting who had a TOL had significantly more sepsis evaluations than did infants who underwent ERCD (23.3 versus 12.5 percent, p=0.0008) and significantly more antibiotic treatments (11.5 versus 4.4 percent, p=0.02).⁹⁰ The absence of predefined criteria for a sepsis evaluation and subsequent antibiotic therapy make interpretation of these results difficult. However, the impact of these interventions are not insignificant as the authors note that an evaluation for sepsis consisted of a complete blood count, blood culture, with lumbar puncture performed at the discretion of the practitioner. Sub analysis of the TOL neonates found increased sepsis evaluations and antibiotic therapy in the TOL group with an epidural compared with TOL neonates without an epidural (29.6 versus 6.0 percent, p=0.0001). Loebel et al also measured “suspected” sepsis, but found no significant differences in neonates born after a TOL versus those delivered by ERCD (3.5 versus 2.7 percent, p= 0.38).⁹⁷

Hook et al similarly reported the incidence of sepsis in neonates who underwent TOL versus those delivered by ERCD. This study was unique in that it was conducted in three hospital settings (Level 1, 2, and 3) and measured the incidence of proven sepsis as well as suspected sepsis.⁷⁹ While suspected sepsis was significantly increased for neonates born by VBAC compared with infants born after a TOL who then delivered by cesarean (12 versus 2 percent, p= 0.001), there were no statistically significant differences between these groups when the outcome was “proven sepsis” (three infants [two percent] versus one infant [0.3 percent], p=NS). Infants without “proven sepsis” still received blood cultures and antibiotic therapy, but were not considered proven to be septic unless a positive blood culture was obtained.

Summary and strength of the evidence on sepsis The overall strength of evidence for the impact of route of delivery upon infant sepsis is low due to imprecise and inconsistent definitions and few studies. While existing studies suggest that there is no significant difference between TOL and ERCD, serious limitations prevent a true understanding of the relationship between route of delivery and sepsis.

Birth Trauma

Two fair quality studies^{90, 93} provided information regarding neonatal trauma. In a large, population based study, Gregory et al found the frequency of trauma to be higher in women who attempted VBAC versus ERCD, regardless of whether the mother had a high-risk clinical condition (3.73 percent attempted VBAC versus 0.77 percent ERCD).⁹³ Subanalyses found that among women who attempt VBAC those with a history of substance abuse were more likely to have neonatal trauma (odds ratio 4.4; 95 percent CI: 1.1 to 18.2) as were those with ruptured membranes longer than 24 hours (odds ratio 4.2; 95 percent CI: 1.7 to 10.2). This study used ICD-9 codes to define trauma (763.1,2,3,4; 7.67.2,3,4,5,6,7,8,9). These codes represented fetal malpositions and varying types of delivery affecting the fetus, as well as skeletal, nerve, and cranial injuries.

In a prospective cohort study comparing neonatal outcome in low-risk women at term undergoing a TOL and low-risk women at term electing a RCD, Fisler et al measured mild bruising (defined as bruising confined to a single extremity) as well as birth injury (not defined).⁹⁰ There was a significant increase in the number of infants who had mild bruising in the TOL group compared with the ERCD group (TOL: 8 percent versus ERCD: 1.5 percent, p=0.0008); however no statistically significant differences were noted between TOL versus ERCD for birth injury (TOL: 1 percent versus ERCD: 1.5 percent, p=0.6). In this study, birth injury was not defined, but the authors noted three infants in the TOL group had cephaohematoma, while two infants in the ERCD group (no labor; scheduled repeat cesarean delivery group) experienced facial nerve palsy.

Summary and strength of evidence on birth trauma The overall strength of evidence for the impact of route of delivery on birth trauma is low largely due to few studies. While existing studies suggest that there a nonsignificant increase in birth trauma for TOL, serious limitations prevent a true understanding of the risk of birth trauma for VBAC compared with ERCD.

Apgar Scores

While Apgar scores suffer from subjectivity and have little long-term predictive value, it is an established and accepted part of the neonatal assessment at the time of delivery. Four good or fair quality cohort studies^{79, 90, 103, 284} all reported no significant difference in 5 minute Apgar scores between infants in TOL groups versus infants in ERCD groups.

Hook et al found no significant differences in an Apgar score of six or less at 5 minutes in infants who underwent a TOL (6/492, 1 percent) versus ERCD (3/497, 1 percent).⁷⁹ Closer examination of the TOL group found significantly fewer neonates had an Apgar score of six or less at 5 minutes if they underwent a VBAC versus when the TOL resulted in a cesarean delivery (26/336, 8 percent versus 22/156, 14 percent, p<0.05). In a fair quality study, Richardson et al similarly found no significant differences in an Apgar score less than seven at 5 minutes in women who underwent a TOL after cesarean delivery versus those who underwent an ERCD (odds ratio 0.5; CI 95 percent 0.2 to 1.2. p= NS).²⁸⁴ However, in a smaller retrospective cohort, fair quality study, Fisler et al found no significant difference in neonates with an Apgar score of less than seven at 5 minutes between neonates who underwent a TOL (regardless of whether they had a VBAC) versus ERCD (1 percent versus 0, p=0.06).⁹⁰ Smith et al, in a large, fair quality population study, found that while the overall incidence of very low Apgar (less than four) at 5 minutes was rare, it occurred more frequently in infants who underwent a TOL (than among those who were delivered by an ERCD (105/15,515 [0.68 percent] versus 40/9014 [0.44 percent] p=0.02).¹⁰³

Summary and strength of evidence on birth trauma. Four studies found no differences in Apgar scores of less than six and seven at 5 minutes in infants undergoing a TOL versus ERCD. Three studies examined the differences in low Apgars (less than seven) at 5 minutes in VBAC versus RCD after a TOL; two of these studies found no difference in Apgar scores of infants born by VBAC versus RCD after a TOL. Future studies that include Apgar score results in their measures of morbidity could be improved by further followup of infants with low Apgar scores as an Apgar score is an intermediate measure of infant health, as well as close attention to classification of exposure to labor and delivery outcome.

Neonatal Intensive Care Unit Admissions

Admission to the NICU is a frequently measured short-term neonatal outcome and has been used as a proxy for serious morbidity. The significance of admission to the NICU can vary by hospital setting, provider experience, provider availability, and pre-established admission criteria. The amount of time a neonate spends in a NICU can vary from a short observation period to a lengthy stay, reflecting the severity of the neonate’s condition.

Eight studies that measured NICU admission met the criteria for review,^{78–80, 90, 97, 156, 284, 286} but none explicated the criteria for NICU admission despite the existence of an American Academy of Pediatrics (AAP) policy statement on levels of neonatal care.²⁸⁹ While admission to a NICU can occur for multiple reasons, it is generally agreed that admission to the NICU results in separation of the neonate from its mother as well as contributing to an increase in the cost of healthcare. Three studies^{78, 80, 156} used reported on the same MFMU cohort. In a fair quality study on risk of uterine rupture in women undergoing TOL, Landon et al isolated term neonates to measure the frequency and likelihood of NICU admission in women who underwent TOL with one previous cesarean delivery compared with women with multiple cesarean deliveries.⁸⁰ Of the 17,898 women in the study, 95 percent had only one prior cesarean; the remainder of women undergoing TOL (N= 975) had two, three, and four prior cesareans. Nine percent of the term neonates whose mothers had one prior cesarean were admitted to the NICU compared with 11.2 percent of the neonates whose mothers had greater than one prior (odds ratio 1.53; 95 percent CI: 1.19 to 1.96, p=0.05); however this study did not find multiple prior cesareans to be predictive of uterine rupture (odds ratio 1.36; 95 percent CI: 0.69 to 2.69, p=0.37). Grobman et al, in a good quality study from the MFMU cohort, examined the effect of IOL on perinatal outcomes in neonates at term with one prior cesarean differentiating between one prior vaginal delivery and no previous vaginal delivery.¹⁵⁶ There were no significant differences in admission to the NICU regardless of type of labor (spontaneous, induced, augmented) or whether women had a pervious vaginal delivery (odds ratio 1.19; 95 percent CI: 0 to 1.47) or no prior vaginal delivery (odds ratio 1.03; 95 percent CI: 0.85 to 1.24). In another study of good quality from the MFMU cohort, Hibbard et al set out to determine whether morbidly obese women have greater maternal and perinatal morbidity with TOL compared with ERCD.⁷⁸ In this secondary analysis she found an increased incidence in the frequency of admission of term neonates to the NICU in women who were obese/morbidly obese (10 percent/13.8 percent) compared with women with a normal BMI (7.4 percent) or who were overweight (7.7 percent, p<0.001). However, neonates of morbidly obese women undergoing TOL were no more likely than neonates of women who had ERCD to experience admission to the NICU (13.8 versus 12.6 percent, odds ratio 1.1; 95 percent CI: 0.9 to 1.3 percent).

The remaining five studies^{79, 90, 97, 284, 286} addressing frequency and likelihood of NICU admission are summarized in Table 35. Two studies attempted to distinguish NICU admission after delivery by VBAC or RCD after TOL. In a study of good quality across all three hospital levels, Hook et al found no significant increase in admissions to a Level 3 NICU in neonates born by TOL compared with ERCD (3 versus 2 percent, p=NS), but found significantly more neonates admitted to NICU if they were born by RCD following a TOL compared with those born by VBAC (7 versus 2 percent, p< 0.007).⁷⁹ Similarly, Kamath et al, in a fair quality study, also found that RCD after TOL increased the likelihood of NICU admission (odds ratio 2.26; 95 percent CI: 0.85 to 6.0, p=0.10).²⁸⁶ However, the greatest likelihood of admission to the NICU (adjusted for multiple covariates of maternal education level, chronic, disease, amniocentesis, choriamnionitis, non-reassuring fetal heart tones, and GA by week) was for term infants who did not experience labor and were born by ERCD (odds ratio 2.93; 95 percent CI: 1.28 to 6.72, p=0.011).

Table 35

Neonatal intensive care unit admissions among term studies.

While costs associated with care vary greatly by setting and region of the U.S., Kamath et al analyzed neonatal length of stay and found neonates born by VBAC stayed 3 days on average compared with 4 days for ERCD (with or without labor) and RCD after TOL (p<0.001).²⁸⁶

Summary and strength of evidence on neonatal intensive care unit admissions. Overall the strength of evidence on the impact of route of delivery on NICU admission is low due to inconsistent and imprecise measures. No studies defined the criteria for admission to the NICU. Six studies found no significant differences in frequency of NICU admissions between TOL and ERCD whereas one reported the greatest risk for NICU admission in infants undergoing an ERCD without labor (odds ratio 2.93) versus a successful VBAC (odds ratio 1.0). Future studies would benefit by description of NICU admission criteria, the reason for admission, and level of support provided to the infant.

Breastfeeding

No studies were found that explored the effect of a TOL versus an ERCD on breastfeeding initiation or continuation.

Impact of the Mode of Delivery on Subsequent Babies

Two fair quality studies were found addressing the impact of a previous cesarean delivery on subsequent unexplained stillbirth in the next pregnancy.^{280, 281} Smith et al, using linked discharge data from a large, population database in Scotland, estimated the risk of antepartum stillbirth in a second pregnancy comparing women with a previous cesarean (N=17,754) to women with no previous cesarean.²⁸¹ The study group included women with a second pregnancy from 1992 to 1998 and excluded fetuses less than 500 grams, deaths due to congenital anomalies, and rhesus isoimmunization. There were 68 stillbirths in the previous cesarean delivery group. After attributed causes for the stillbirths (e.g., toxemia, hemorrhage, mechanical, maternal issues) were noted, there were more unexplained stillbirths found in the previous cesarean delivery group (43 unexplained stillbirths out of 17,754 births or 0.24 percent of births) compared with women without a prior scar (163 unexplained stillbirths out of 102,879 births or 0.16 percent of births). Adjusting for maternal and demographic characteristics did not lessen the association between previous cesarean and unexplained stillbirth. Smith also used time-to-event analyses to predict the prospective risk of a stillbirth at 39 weeks or greater gestation to be 0.11 percent of women compared with 0.05 percent of women without a prior cesarean delivery. The authors noted that despite a wealth of clinical and demographic data in the dataset they did not have access to maternal weight, and therefore were unable to adjust for BMI and obesity as a maternal risk factor.

In another large U.S. dataset, Bahtiyar et al conducted a cross-sectional study of singleton, term pregnancies from 1995 to 1997 in women with a prior cesarean delivery compared with women without a prior cesarean delivery, excluding any underlying maternal or fetal abnormality.²⁸⁰ Women were stratified in order to detect the study group of interest and to reduce potential confounders, with the final group of women with only one prior delivery (cesarean versus vaginal birth). Rate of stillbirth in the subsequent pregnancy was compared between the two groups. While not statistically significant, there were fewer stillbirths in the previous cesarean group compared with women without a prior cesarean in the second pregnancy (0.072 percent of births compared with 0.080 percent of births, RR 0.90; 95 percent CI: 0.076 to 0.106 percent). While this study is strengthened by its large cohort size (1.7 million for comparison in the study group described), it was difficult to determine the exact number of cases in each current pregnancy group. The authors did not attempt to classify stillbirths by cause and based study data on birth and death certificates.

Summary of mode of delivery on subsequent babies. Two studies were reviewed to determine the risk of stillbirth in subsequent pregnancies in women with a prior cesarean delivery. These studies produced conflicting results with one study showing that prior cesarean increases the risk for unexplained stillbirth in next pregnancy and the other study showing no difference in risk for stillbirth in the next pregnancy. One study included early gestations, while the other study was limited to term gestations. Both studies are limited by their retrospective design and relied on large perinatal databases while employing various methodologies to overcome confounding.

Neurological Development

No studies were found that measured the impact of a TOL versus an ERCD on neonatal neurological development. In order to examine the neurological development, studies that extend beyond the immediate postpartum period would be required.

Special Considerations