Chapter  133:  Cesarean Delivery on Maternal Request

KQ1: Incidence and Trends of Cesarean Delivery on Maternal Request

KQ 1 referred to the incidence and trends in cesarean deliveries over time in developed countries; it made specific reference to primary cesarean before onset of labor, CDMR, medical indications, and malpresentation as proportions of total cesarean deliveries. The absence of data to answer this question is striking. Regarding incidence, the available literature yielded rates of cesarean deliveries as a proportion of all deliveries for a wide array of time points and countries. For 2001 in the United States, data suggest rates of more than 25 percent. Elsewhere in the developed world for 2001, rates of cesarean delivery ranged from 14 percent in The Netherlands to 35 percent in Italy. Since 2001, the rates of cesarean delivery have risen in the United States; recent figures put the rate at more than 29 percent for 2004.

The rate of cesarean deliveries is rising worldwide. Both “elective” cesarean deliveries (sometimes defined as unlabored) and “nonelective” cesarean deliveries contribute to this rise; however, the proportions vary by country, study, and time period. Four studies distinguished between prelabor primary and repeat cesareans. An Irish study reported an unlabored primary cesarean delivery rate of 18.9 percent of all cesarean deliveries during the 12-year period from 1989 to 2000. One study in Australia showed that prelabor primary cesarean delivery as a percentage of all deliveries rose from 4.1 percent in 1980 to 4.8 percent in 1987. In the United States, primary prelabor cesarean delivery rates were approximately 5 percent of all deliveries in 1996 and approximately 7 percent in 2001. In 2001, “primary elective” prelabor cesarean rate as a proportion of all cesarean deliveries was 28.3 percent in the United States.

The extent to which CDMR is contributing to the rise in cesareans remains unclear. Finally, we did not find sufficient data to comment on medical indications or malpresentation as a proportion of all cesarean deliveries.

KQ 2: Outcomes of Cesarean Delivery on Maternal Request

Overall, few moderately relevant studies were available, and the strength of evidence is weak for nearly all outcomes.

Maternal outcomes for primary cesarean deliveries. Mortality. Four studies suggested no evidence of difference in maternal mortality associated with planned vaginal versus planned cesarean delivery. These studies provided weak evidence overall.

Infection. The 12 studies that included maternal infection as an outcome provided weak evidence that the risk of maternal infection was lower with planned cesarean than with unplanned cesarean delivery and lower for vaginal than for cesarean delivery.

Anesthetic complications. Two studies showed a lower rate of anesthetic complications with planned vaginal than with planned cesarean delivery; the third reported no significant difference between these two routes of delivery. These studies provided weak evidence suggesting a lower rate of anesthetic complications with planned vaginal delivery.

Hemorrhage and blood transfusion. Eleven studies provided moderate strength of evidence showing a lower risk of hemorrhage and blood transfusion in planned cesareans than in vaginal delivery. These studies also yielded evidence of lower hemorrhage or blood transfusion in planned cesareans than in unplanned cesareans.

Hysterectomy. Three studies yielded weak evidence on the association between emergency hysterectomy after childbirth and either planned vaginal or planned cesarean delivery. The rarity of the outcome results in insufficient statistical power to draw firm conclusions regarding the risk associated with either delivery route.

Thromboembolism. Eight studies provided weak evidence for an association between thromboembolism and planned vaginal or planned cesarean delivery. Studies reported no consistent direction or magnitude of effect.

Surgical complications. Ten studies provided weak evidence on surgical complications associated with planned vaginal and planned cesarean delivery. Studies generally showed a lower risk of surgical complications in planned “elective” cesarean than unplanned “emergency” or “labored” cesarean deliveries.

Breastfeeding. One study provided weak evidence that although women with planned vaginal deliveries may initiate breastfeeding sooner than women with planned cesarean deliveries, they do not report any difference in the duration of breastfeeding. Other evidence suggests that women are more likely to bottlefeed following a cesarean delivery (planned or unplanned) compared with a vaginal delivery.

Postpartum pain. Four articles (from three studies) reported on postpartum pain using various pain measures at different time periods. Together, these studies provide weak evidence of no significant difference in pain between modes of delivery, but they draw from populations with breech deliveries and may, therefore, overestimate the pain in the planned vaginal delivery group.

Psychological outcomes: postpartum depression. Two studies provide weak evidence suggesting no differences in postpartum depression by delivery route. As with pain, studies with breech populations likely overestimated the rate of complications, interventions, and possible negative psychological outcomes in the planned vaginal delivery group.

Psychological outcomes: other. Seven articles (from six studies) yielded weak evidence about a range of other psychological outcomes. The data were consistent in reporting that women who had an unplanned cesarean birth or an instrumental vaginal delivery were more likely to experience adverse psychological outcomes than were women who either underwent a spontaneous vaginal or a planned cesarean birth. The variety of outcomes and measures makes a summative assessment of other outcomes challenging.

Maternal length of stay. Four studies provided moderate evidence that length of stay is higher for cesarean delivery, planned or otherwise, than for vaginal delivery.

Urinary incontinence. Nine articles (from eight studies) provided weak evidence that rates of stress urinary incontinence for planned “elective” cesarean section were either lower than or no different from those for vaginal delivery. Numerous problems limit evidence on this outcome: lack of high-quality prospective studies that compare planned routes of delivery, have adequate power, include comprehensive long-term followup, account for multiple deliveries, account for variations in practice patterns including use of epidural anesthesia and episiotomy, use validated urinary questionnaires administered at consistent time points from delivery, and define incontinence in a standardized fashion by its occurrence, severity, and impact on quality of life.

Anorectal function. Seven articles (from six studies) provided weak evidence showing a reduced risk of anal incontinence in planned cesarean deliveries compared with unplanned cesarean or instrumental vaginal deliveries. Evidence was inconsistent about differences between planned cesarean and spontaneous vaginal delivery.

Pelvic organ prolapse. We found no evidence on the association between pelvic organ prolapse and planned vaginal or planned cesarean delivery.

Sexual function. One study provided weak evidence that sexual function does not differ by planned route of delivery.

Maternal outcomes relevant to subsequent cesarean delivery. Subsequent fertility issues. Studies not included in this review suggests a higher risk with all cesarean deliveries (unplanned or planned), but we found no reliable evidence of difference relevant to CDMR.

Subsequent uterine rupture. A recent update of a systematic review on the outcomes of vaginal birth after cesarean (VBAC) provided moderate evidence on subsequent uterine rupture. The update found no statistically significant differences between trial of labor after cesarean and elective repeat cesarean delivery with regard to rates of asymptomatic uterine rupture rates. The update noted that two studies of fair or good quality found a small but higher risk of symptomatic uterine rupture in trial of labor after cesarean than in elective repeat cesarean delivery.

Placenta previa. Given that placenta previa is the most common placental implantation anomaly, we updated a recent meta-analysis examining the relationship between placenta previa and a history of cesarean delivery. Our update supports the earlier meta-analytic conclusion that the odds of placenta previa are associated with advancing maternal age and increasing parity. The literature provided moderate evidence that the risk of placenta previa increases with previous cesarean delivery.

Subsequent stillbirth. Studies not included in this review suggest a higher risk with all cesarean deliveries (unplanned or planned), but we found no reliable evidence of difference relevant to CDMR.

Neonatal outcomes. Fetal mortality. We found no studies that addressed fetal (in utero) deaths.

Neonatal mortality. Two studies provided weak evidence on neonatal mortality. The studies suggested a higher risk for all cesareans (planned or unplanned) than for spontaneous vaginal delivery. The studies did not control for underlying maternal or neonatal indications for cesarean or were underpowered for such a rare outcome, leading to limited ability to draw conclusions on this outcome.

Unexpected (iatrogenic) prematurity. We found no study that addressed unexpected prematurity and allowed comparisons by type of cesarean with intended or actual vaginal birth.

Respiratory morbidity. Measures of respiratory morbidity range from transient tachypnea of the newborn (TTN) to severe respiratory distress syndrome (RDS) with long-term sequelae. Nine articles yielded moderate evidence that the risk of variably defined “respiratory morbidity” was higher for all cesarean births than for vaginal deliveries. This risk reduces with advancing gestational age. Studies did not assess meconium aspiration syndrome by mode of delivery.

Transition issues. One study reported on this outcome, but the significant issues of appropriate categorization in this study make interpreting the data difficult. We consider the available evidence insufficient to judge the direction of effect.

Neonatal asphyxia or encephalopathy. Two studies provided weak evidence of a higher risk of neonatal encephalopathy associated with operative vaginal deliveries and “emergency” or “labored” cesareans than with spontaneous vaginal delivery.

Intracranial hemorrhage. One study provided weak evidence on intracranial (subdural/cerebral, intraventricular, and subarachnoid) hemorrhage. The prelabor cesarean deliveries included those done for maternal or neonatal indications, so they likely involved cesareans for placenta previa and fetal anomalies, which may independently increase the risk of intracranial hemorrhage. Despite the higher theoretical risk for prelabor cesarean deliveries, this study did not find any significant difference between spontaneous vaginal delivery and prelabor cesarean deliveries. It did show consistently higher rates of intracranial hemorrhage for assisted vaginal deliveries and cesarean deliveries in labor.

Facial nerve injury. One study provided weak evidence that the risk of facial nerve injury varies by mode of delivery; the risk is higher for forceps and the combined use of forceps and vacuum delivery than for spontaneous vaginal delivery. These findings suggested that CDMR posed no risk for facial nerve injury greater than that associated with planned vaginal delivery.

Brachial plexus injury. One study provides weak evidence that the incidence of brachial plexus injury is lower in cesarean delivery than in vaginal delivery.

Fetal laceration. Two studies provided weak evidence on fetal lacerations based on data limited to cesarean deliveries. They reported a higher rate of fetal lacerations among emergency and labored cesarean than among elective cesarean delivery.

Neonatal length of hospital stay. One study provided weak evidence that the neonatal length of hospital stay is higher for “elective” cesarean delivery than for vaginal delivery.

Long-term neonatal outcomes. We did not find any evidence on long-term neonatal outcomes.

KQ 3: Modifiers of Cesarean Delivery on Maternal Request

The evidence on effect modifiers is sparse and pertains to only a few outcomes for KQ 2. Five studies provided evidence on the modifiers of CDMR, specifically neonatal respiratory distress, infectious morbidity, and urinary incontinence.

With regard to respiratory morbidity, results showed a consistent decrease in respiratory morbidity as gestational age rises, despite differences in inclusion criteria and definitions of elective cesarean delivery. Gestational age appears to play a lesser role as a risk factor for fetal respiratory distress in planned vaginal delivery than in planned cesarean.

With regard to infectious morbidity, the single study we found suggested no effect of physician experience, incision type, maternal age, or prophylactic antibiotics on infectious morbidity; it did suggest that the risk was higher among obese or black patients than among other women. Pelvic floor exercises decreased the risk of urinary incontinence; pre-pregnancy body mass index increased it.

Given the lack of evidence directly comparing effect modifiers in a population with planned CDMR with those in a population with planned vaginal delivery, inferences about effect modifiers must be drawn cautiously. Furthermore, most studies did not adjust for confounders, so results must be interpreted as crude estimates.

KQ 1: Trend and Incidence of Cesarean Delivery

This question includes trend and incidence of cesarean delivery over time and covers the contribution of primary prelabor cesarean deliveries. KQ 1 further seeks to determine the contribution of CDMR and other cesareans for other indications such as repeat elective cesarean deliveries, unlabored cesareans for medical indications (maternal or neonatal), unlabored cesareans for malpresentation, and labored cesareans after a planned vaginal delivery (Figure 3). These groupings will be particularly relevant to both maternal and neonatal outcomes because short- and long-term risks may be associated with the degree of effort exerted to achieve a vaginal delivery.¹³

Despite the theoretical importance of these distinctions, obtaining accurate data on the rate of cesarean delivery truly on maternal request in the absence of maternal or neonatal indications is challenging. Determining the true prevalence of CDMR in this country is difficult because such deliveries are often coded with other indications, possibly reflecting insurance coverage issues.⁴ However, some evidence suggests that such deliveries do occur and possibly at an increasing rate.^{5, 14}

We queried other sources of data to answer this key question. These include CDC, National Vital and Health Statistics, the World Health Organization, and sources from other nations such as Statistics Canada, the Australian Department of Human Services, and the United Kingdom Department of Health.

We defined “developed countries” as the United States, Canada, the United Kingdom, Western Europe, Israel, Australia, New Zealand, and Japan. We also tracked citations from other countries, such as Brazil.

KQ 2: Effect of Planned Route of Delivery on Outcomes

Several factors make interpretation of the available evidence challenging. We summarize them in the following section.

Comparison of planned routes of delivery (intent to treat). As explained above, the appropriate comparison is that of intent: planned vaginal delivery compared with planned CDMR. The majority of studies included in this systematic review report outcomes by actual route of delivery. A design centered on actual delivery route often allows investigators to distinguish between labored and unlabored cesarean deliveries. In studies limited to unlabored cesareans, women who present in labor before their scheduled date of delivery are, by definition, excluded. Excluding these women may overestimate potential benefits (e.g., reduction in pelvic floor disorders) and potential harms (e.g., neonatal respiratory morbidity) associated with CDMR because the studies then cannot account for any effect that labor has on outcomes of interest.¹⁵–¹⁷ Studies that include both labored and unlabored planned cesareans may have a rate of labor that exceeds the rate of labor expected for a population planning CDMR and may allow for a longer period of time in labor before cesarean delivery.

Appropriate proxies for CDMR. We expected high-quality data on CDMR per se to be limited because CDMR is rarely listed as an indication for a cesarean delivery Available studies include a wide range of indications for cesarean that are highly variable in their relevance to CDMR. Use of cesarean for breech or other malpresentations is currently the closest proxy for CDMR. However, studies of cesarean deliveries for breech were designed with neonatal outcomes as primary endpoints; therefore, they may be limited in their ability to serve as proxies for CDMR for maternal outcomes. For instance, the International Term Breech trial (hereafter Breech Trial) allowed patients who presented in labor to be randomized to a cesarean delivery¹⁸ without adjusting for the length of the labor before cesarean or the length of time the membranes had been ruptured. As indicated above, the ideal comparison would involve intent and compare planned vaginal with planned CDMR. Thus, any protocol for CDMR would have some women going into labor before their planned cesarean; data from such deliveries ought to be analyzed as part of the planned CDMR group. However, the extent to which cesarean for breech serves as an accurate surrogate for CDMR is unclear because of uncertainty as to whether the time period between presentation in labor and cesarean delivery for breech is similar to that of CDMR. Issues such as prolonged rupture of membranes before the decision to perform a cesarean may increase the risk of other complications, such as maternal and neonatal infections and length of hospital stay.

Another major limitation to the use of breech as a proxy for CDMR is that when comparing study groups based on intent to treat, the risk of requiring a cesarean in the planned vaginal delivery group is likely to be significantly higher than if the fetus were vertex (head first).

Changing practice patterns. Practice patterns have changed considerably for both cesarean and vaginal deliveries over the past two decades. Historically, quantifying the risks and benefits of vaginal and cesarean births has been difficult, as the data on the risks associated with cesarean were based on older studies when cesarean deliveries were routinely performed under general anesthesia, after prolonged labor, and without the benefit of prophylactic antibiotics and thromboprophylaxis. Such surgical procedures are not comparable to CDMR under current standard practices. The absolute risks associated with a planned CDMR are likely to be lower in today's environment than they were previously,¹⁹–²³ and they are dropping.²⁴ Similarly, clinical management of planned vaginal delivery has also been improving, as in the declining use of episiotomy.^{25, 26}

The TEP and the SOS Conference panel chair decided to exclude studies published before 1990 to focus the systematic review on studies with practice patterns similar to contemporary norms. However, the studies we examine in this review do not necessarily include “best practices” for either vaginal or cesarean routes of delivery and demonstrate variable practice patterns among providers. For instance, studies generally do not clarify whether prophylactic antibiotics were administered for cesarean; this step, of course, can affect rates of maternal infection. Similarly, studies of vaginal delivery may reflect overuse of episiotomy, inappropriate thresholds for performing cesarean delivery, and inadequate management of labor. Therefore, a comparison of planned vaginal and planned cesarean ought to include the best clinical practice patterns for each of these intended routes of delivery. In the absence of information on the extent to which studies deviate from ideal practice patterns, how the balance of harms and benefits may shift in an ideal practice environment remains unclear.

Confounders. As noted earlier, ideal comparisons include planned vaginal delivery vs. planned cesarean delivery. The comparison of actual routes of delivery may result in inadequate assessment of confounders that influence both route of delivery and maternal or neonatal outcomes. For instance, confounders such as multiple gestations, placenta previa, and polyhydramnios (excess amniotic fluid) may increase the likelihood of preterm labor and delivery, may influence the recommended route of delivery, and may also result in poor maternal and neonatal outcomes.

Statistical power. The consequences of a fundamental shift to higher rates of CDMR are profound. They should be examined in well-designed studies that are adequately powered to detect clinically meaningful differences. The available literature that we discuss generally has sample sizes lower than are necessary to achieve adequate power, especially for rare outcomes.

Appropriate timing of outcome measurements. Decisions made in the delivery room have lifelong implications for the mother and infant. Ideal studies require that outcomes be assessed at time periods appropriate for that particular end result of care. Studies of the association between parturition-related variables including routes of delivery and pelvic floor disorders (urinary incontinence, pelvic organ prolapse, anal incontinence, and sexual dysfunction) are often limited to the immediate postpartum period. Assessing whether the condition results in long-term impairment is difficult in these studies.

Long-term outcome studies, although relatively uncommon, are often retrospective in design; they draw associations between current pelvic floor complaints and previous obstetrical events, sometimes decades earlier. These studies are often unable to collect specific information regarding planned routes of delivery or even sufficient detail regarding actual routes of delivery. Therefore, they frequently are unable to control for many important confounders such as interval pregnancies and deliveries and other factors that have been implicated in the development of pelvic floor disorders, such as length of labor, use of vacuum or forceps, obesity, smoking, constipation or chronic straining, or previous reconstructive pelvic surgery.

Measurement of outcomes (comprehensiveness, severity, and utility). Ideally, a systematic review of the outcomes of planned route of delivery should provide a comprehensive assessment of outcomes, accounting for the severity of symptoms and the utility of various outcomes to patients. For instance, accurate measurement of neonatal respiratory morbidity should include the risks of all forms of harm associated with planned route of delivery, including potentially higher risks of meconium aspiration in planned vaginal deliveries²⁷ and potentially higher risks of transient tachypnea of the newborn and respiratory distress syndrome in planned cesarean deliveries.

The issue of severity rating is particularly important for pelvic floor outcomes such as urinary incontinence, pelvic organ prolapse, or anal incontinence. An undifferentiated measure of urinary incontinence that does not account for severity would mask the considerable difference in quality of life between a small amount of leakage that occurs rarely and severe, daily urinary leakage. Similarly, neonatal outcomes such as respiratory morbidity need to be categorized and analyzed by degree of severity. For instance, transient tachypnea of the newborn (TTN) and respiratory distress syndrome (RDS) represent extremes of severity; investigators should not group them into a single measure of respiratory morbidity, because doing so may obscure meaningful differences between groups.

Factoring in both severity and utility when assessing the overall benefit and harm of CDMR is critically important. A woman considering a planned route of cesarean delivery needs to assess comprehensively both short- and long-term risks, to both herself and her infant, and in both the current pregnancy and future pregnancies. Currently, clinicians and others have little or no way to judge the “priority” of a range of possible outcomes. For instance, urinary incontinence needs to be described in a manner that relates both its occurrence and severity and that provides a utility weighting relative to other potential outcomes such as wound infection. Similarly, in assessing overall harms and benefits to the neonate, the potentially higher risk of neonatal respiratory morbidity (TTN and RDS) associated with a planned CDMR needs to be weighed against the potential reduction in the rate of other outcomes such as stillbirths after 39 weeks, intrapartum deaths, and shoulder dystocias (emergency occurring when infant's shoulder gets “stuck”) associated with a planned vaginal delivery.

KQ 3: Magnitude of Benefits and Harms

As suggested by our analytic framework (Figure 3), the choice of CDMR or a planned vaginal delivery, as well as the rates and severity of subsequent maternal and fetal or neonatal benefits and harms, may be influenced by several maternal, fetal, provider, and health care system characteristics. Given that few studies control for or assess the effect of such factors on maternal and fetal outcomes, we acknowledge the challenges of measuring the magnitude of these factors on maternal and fetal or neonatal outcomes. As in KQ 2, we stratified the analysis into unlabored and labored cesarean deliveries as well as planned vaginal deliveries.

KQ 4: Future Research

We anticipate that even after this comprehensive systematic review has been discussed at the SOS conference and published, appreciable uncertainty will remain about the risks associated with CDMR. Suggestions to address these limitations of the literature, put forward in our discussion of KQ 4, will guide the development and direction of future research.

Understanding some of the gaps in the literature at this point may help readers interpret our analyses and findings for KQs 1, 2, and 3. Especially important are problems related to the characterization of CDMR and other modes of delivery that typically serve as proxies for CDMR. These include cesareans for breech and other ambiguous categorizations, which may be called “elective,” “planned,” “nonemergency,” “unlabored,” and “scheduled” cesareans. This variability in language is not trivial, and readers of this evidence report are cautioned against assuming that various research teams in fact mean the same thing by the same term or that use of different terms accurately depicts different situations.

In addition, we have assessed maternal and neonatal outcomes and weighted them based on the level of relevance to CDMR, the quality of individual study, and the overall strength of evidence for each particular outcome measure. In taking this approach, we have identified several outcomes that require substantial additional research. The analytic framework provides the infrastructure for designing future studies by highlighting particularly relevant comparison groups. Although a comprehensive assessment that balances outcomes based on their relative rates and utilities or disutilities is the ideal, this goal is probably still unattainable.

A Note on Terminology

The purpose of this systematic review is to address CDMR, but the paucity of literature on the topic requires the use of proxies. As noted above, studies in this review use nonstandardized terms such as “elective,” “scheduled,” “planned,” “labored,” “unlabored,” “urgent,” “emergent,” and “emergency” cesarean deliveries. Table 2 provides a glossary of these general terms as they are commonly used. Not all studies use these exact definitions; in fact, they may use definitions that differ significantly from those employed in other studies. For instance, the term “primary elective” cesarean is widely interpreted as referring to a woman's first cesarean delivery, planned for a wide range of maternal and fetal indications and generally distinguished from emergency cesarean delivery and labored cesarean delivery after planned vaginal delivery. However, one study included 9 percent of women undergoing a repeat cesarean in their “primary elective” cesarean group. The authors of this study defined the term “primary elective” cesarean as a “planned operation in which the patient had been admitted to the hospital at least 8 hours before the cesarean without symptoms of labor,”^{28 p. 2} allowing for inclusion of repeat cesareans.

In discussing studies, we clearly specify each study's definition with respect to medical indications for cesarean and laboring status. In recognition of the variation across definitions, we elected to use the term that the authors used, denoting such terms by using quotation marks, rather than try to impose a single, overarching term such as elective cesarean delivery or CDMR.

Inclusion and Exclusion Criteria

Our inclusion and exclusion criteria are documented in Table 3. As noted in Chapter 1, we limited our searches to articles published in or after 1990 because of the significant advances in operative techniques, anesthesia, availability of antibiotics, and neonatal care over the past several decades that have resulted in a decline in maternal and neonatal mortality.¹ We also restricted our searches to developed countries so that we could have comparable data on the standard of care. Based on recommendations from the Technical Expert Panel (TEP), we tracked citations from Brazil, which has long been documented to have high rates of cesarean deliveries;^{29, 30} however, no study from Brazil met our inclusion criteria.

Because our searches focused on the comparison of planned cesarean delivery to planned vaginal delivery, we recognized that we were unlikely to capture relevant studies on placental implantation abnormalities. On the advice of our TEP, our summary of this topic consists of an update of a recent systematic review on placenta previa.³¹ Because of time and resource limitations, however, we could not address other placental implantation abnormalities such as accreta (abnormally firm attachment of the placenta to the uterine wall) and percreta (extension of the placenta through the entire wall of the uterus) that may also be associated with a history of cesarean deliveries.

Similarly, our search strategy focused on primary cesarean deliveries, excluding studies limited to repeat cesarean deliveries. For that reason, we could not capture studies that examined outcomes such as uterine rupture related to subsequent deliveries in women with prior cesarean deliveries. Again on the advice of our TEP, we address this important topic using information from an update of a recent review on vaginal birth after cesarean (VBAC).³²

We excluded studies that (1) did not report on women of reproductive age; (2) were published in languages other than English (given the available time and resources); (3) did not report information pertinent to the key clinical questions; (4) had fewer than 50 subjects for randomized controlled trials (RCTs) or 100 subjects for observational studies; and (5) were not original studies. Additionally, and in consultation with the TEP, we excluded studies that did not include data on both planned cesarean delivery and planned vaginal delivery for KQ 1 and KQ 2. As a consequence of this search strategy, we cannot address outcomes from vaginal delivery alone, without reference to a cesarean delivery comparison group. A review of the outcomes from vaginal delivery alone was beyond the scope and resources available. As a consequence of this limitation, we are not able to address modifiers of vaginal deliveries alone in KQ 3.

We also excluded studies that reported on an undefined group of cesarean deliveries; many of the maternal and neonatal indications that would have been included were so highly associated with significant morbidity as to preclude any meaningful extrapolation to CDMR.

Literature Search and Retrieval Process

Databases. We used multifaceted search strategies to include current and valid research on the key questions. We used standard electronic databases: MEDLINE^®, Cochrane Collaboration resources, and Embase. We also hand-searched the reference lists of relevant articles to make sure that we did not missing any relevant studies. We consulted with the TEP about any studies or trials that are currently under way or that may not be published yet.

Search terms. Based on the inclusion/exclusion criteria above, we generated a list of Medical Subject Heading (MeSH) search terms (Appendix A).^* Our TEP also reviewed these terms to ensure that we were not missing any critical areas, and this list represents our collective decisions as to the MeSH terms used for all searches. We needed to conduct several focused searches to capture a wide pool of relevant studies because the term “cesarean delivery on maternal request” is not a standard indexing term (Table 4). The SOS Conference panel had set priorities for outcomes of interest, and those guided our selection of the specific outcomes for this review. Our initial searches did not capture key citations dealing with neonatal outcomes; based on the advice of the TEP, we conducted additional searches to capture relevant citations.

Figure 3 presents the yield and results from our search, which we conducted from April through June 2005. Beginning with a yield of 1,402 articles, we retained 65 articles comprising 62 studies that we determined were relevant to address our key questions and met our inclusion/exclusion criteria (Figure 4). Peer reviewers suggested several additional citations, of which 4 articles comprised 3 new studies that did not duplicate ones we had already identified and excluded. We reviewed titles and abstracts of the remaining suggestions against the basic inclusion criteria above; we retained relevant articles, all published after our search cutoff date, and used them as appropriate in the discussion in Chapter 4.

Article selection process. Once we had identified articles through the electronic database searches, review articles, and bibliographies, we examined abstracts of articles to determine whether studies met our criteria. Two reviewers separately evaluated the abstracts for inclusion or exclusion, using an Abstract Review Form (see Appendix B). If one abstractor concluded that the article should be included in the review, we retained it. The group included four physicians—Anthony Visco, MD (Scientific Director); Katherine Hartmann, MD, PhD (Senior Advisor); Jennifer Wu, MD; and Gerald Gartlehner, MD, MPH (Study Coordinator). It also included one health services researcher—Meera Viswanathan, PhD (Study Director) and three epidemiologists—Michele Jonsson Funk, PhD, Rachel Palmieri, BS, and Shauna Hay, BS.

Four hundred and ninety articles required full review because of missing or uninformative abstracts. For the full article review, one reviewer read each article and decided whether it met our inclusion criteria, using a Full Text Inclusion/Exclusion Form (Appendix B). Articles excluded at the full-article review stage and reasons for their exclusion are listed in Appendix D.

Development of Evidence Tables and Data Abstraction Process

The staff members who conducted this systematic review jointly developed the data abstraction tables (Appendix B ^†) and evidence tables (Appendix C). We designed the tables to provide sufficient information to enable readers to understand the studies and to determine their quality; we gave particular emphasis to essential information related to our key questions. The format of the evidence tables was based on successful designs used for prior systematic reviews.

We trained abstractors in entering data into the tables by having them abstract several articles and then reconvening as a group to discuss the utility of the table design. The abstractors repeated this process through several iterations until they decided that the tables included the appropriate categories for gathering the information contained in the articles.

All team members shared the task of initially entering information into the data abstraction forms. Another member of the team also reviewed the articles and edited all initial table entries for accuracy, completeness, and consistency. The two abstractors reconciled all disagreements concerning the information reported in the abstraction forms. The full research team met regularly during the article abstraction period and discussed global issues related to the data abstraction process.

We then entered the data from the abstraction forms into evidence tables and once again checked for consistency and accuracy.

The final evidence tables are presented in their entirety in Appendix C. Studies are presented in the evidence tables alphabetically with the last name of the first author. A list of abbreviations and acronyms used in the tables appears at the beginning of that appendix.

Quality and Strength of Evidence Evaluation

Rating the relevance of individual articles. The vast majority of studies reported results on actual route of delivery rather than planned route of delivery (intent to treat), which led to the limitations introduced in Chapter 1. Initial review of the literature demonstrated several ambiguous definitions that presented the authors of this review with several challenges. The use of the phrase “elective cesarean delivery” was particularly problematic because of its wide range of definitions. Some investigators used the term “elective” to refer to situations in which a vaginal delivery was contraindicated, such as with placenta previa. Others used it to describe situations in which vaginal delivery could have been attempted, such as with breech presentation, active herpes simplex virus, or repeat cesarean delivery. Still others used it but failed to define it further, precluding reviewers from understanding either the labor status or the indications for the cesarean.

We found it impossible to arrive at any meaningful summary of the literature without explicitly addressing the issues of how to characterize the groups in these studies and, thus, how to categorize the studies themselves. To address this ambiguity, we developed a four-tier classification system of relevance to CDMR based on the following criteria: (1) whether studies analyzed outcomes by planned route of delivery (trials of route of delivery); (2) whether CDMR was included as a comparison group (high relevance); (3) whether comparison groups comprised planned cesareans (moderate relevance), and (4) whether studies involved undefined “elective” or a mix of planned and unplanned, unlabored cesareans (low relevance).

Table 5 presents relevance ratings for all studies included in this review. The first three categories of relevance—high (H), moderate (M), and low (L)—are explained below.

Because we view the first criterion above as the ideal comparison, we assigned trials of routes of delivery,^‡ comparing prospectively assigned planned routes of delivery for breech presentation, a relevance rating of “T” to distinguish them from other studies that dealt with actual delivery routes. However, we note that trials of route of delivery for breech presentation are limited in their relevance to CDMR for primarily vertex (head first) presentation for four main reasons: (1) they cannot be included in a summary of neonatal outcomes because the confounding effect of breech presentation in the sample of women could negatively influence neonatal outcomes; (2) the extent to which cesarean for breech serves as an accurate surrogate for CDMR is unclear because of uncertainty as to whether the time period between presentation in labor and cesarean delivery for breech is similar to that for CDMR; (3) the risk that a cesarean would be performed in the planned vaginal breech delivery group is likely to be significantly higher than if the fetus were vertex; and (4) the inclusion of multiparous patients, including some with previous cesarean deliveries, results in significantly different risks and benefits than for the central focus of this review, namely, primary CDMR.

Many studies included a combination of planned and unplanned, labored and unlabored cesarean births for maternal or neonatal indications in an “elective” cesarean group. Using our relevance classification scheme, we sorted this range of studies into groups of literature with high, moderate, and low relevance to CDMR. Studies with a cesarean delivery group performed solely on maternal request, without any maternal or neonatal indications, were considered the most highly relevant to the central question of this systematic review; we assigned them a relevance rating of A. As expected, however, we found no such published studies for KQ 2 and KQ 3 and only one such study for KQ 1.

Studies in the moderately relevant category were all planned cesareans, but they included labored, unlabored, or a mix of labored and unlabored cesarean deliveries. Some studies included cesareans planned and performed, before labor, for maternal or neonatal indications (or both). Such studies may understate the risk of CDMR, because an accurate assessment of outcomes should include both labored and unlabored cesareans when comparison groups are separated by intent; that is, planned CDMR versus planned vaginal delivery. The group also included studies involving planned and performed cesareans for maternal and/or neonatal indications but with a mix of labored and unlabored deliveries. These may overestimate the risk of planned CDMR if the rate of labored cesareans in the study exceeds the rate of labor before scheduled delivery in a population considering CDMR.

Studies of low relevance did not define the “elective” cesarean delivery group, or they included a mix of planned and unplanned unlabored cesarean deliveries. The chief uncertainty concerns the degree to which the “elective” cesarean delivery group included emergency or labored cesareans. Emergency indications that would potentially be included in such a category include abruption, maternal trauma, and fetal distress; each could increase maternal or neonatal morbidity considerably.

Rating the quality of individual articles. We developed our approach to assessing the quality of individual articles (see Appendix B for Quality Rating Forms) based on the domains and elements for RCTs and nonrandomized observational studies recommended in the evidence report by West and colleagues, Systems to Rate the Strength of Scientific Evidence.⁸⁵

We also elected to limit our quality ratings to studies with at least moderate relevance because low-relevance studies were generally designed for purposes other than addressing planned cesarean delivery.

The only study eligible to be rated with an RCT form was the International Term Breech Trial (hereafter Breech Trial).^{18, 20, 83} Two of the criteria listed below (randomization approach, post-randomization exclusions) apply to the entire study; all others (masking, operational definitions and measures, loss to followup, and statistical analysis) apply to each article individually. We elected to rate each article in the Breech Trial individually because of significant variations in the article-specific criteria.

• 1

  Randomization approach and implementation: This item judged whether the approach described a valid method of randomization, whether allocation concealment was achieved, and whether balance was documented across study groups.

  Approach: If the study assigned the groups in a manner inconsistent with true randomization methods, it had the potential to automatically receive a poor rating for this category and overall. If the study had merely stated that if “randomly assigned” the groups and either had no balance or did not report on balance, it would have received a poor rating. A study with no documentation of concealment or with inadequate concealment methods would have been rated poor if the study had poor balance of allocation or if balance was not documented. A study with potentially poor concealment would have been rated fair if they documented good balance.

• 2

  Post-randomization exclusions: This item captured how many post-randomization exclusions were explicitly stated.

  Approach: In typical randomized trials, intention-to-treat analysis is expected. Any exclusions after randomization would have been considered inappropriate and would have led to a rating of poor.

• 3

  Masking: This item was relevant only to outcome assessors.

  Approach: If the outcome assessors were adequately masked within the possibilities of the study design, we rated the category as good. If there was a mix of masking among the outcomes, we rated the category as fair. If masking was not done at all and not attempted, we rated the category as poor.

• 4

  Operational definitions and measurements: This item judged the quality of the operational definitions of the outcomes (i.e., were they adequately described) and whether they were adequately collected (i.e., was the method sufficient and appropriate).

  Approach: We rated this category on the basis of an average across all outcomes for each timepoint and the ability to define and measure them. Good definitions and measurement include the following: validated questionnaires, detailed time points in question, details about what was asked of the patient, medical chart abstractions, and clinical examination or assessment. Failure to use such methods resulted in a rating in the fair-to-poor range, depending on how the article collected the information.

• 5

  Loss to followup: This item collected percentages of followup at every time point in the study at which data were collected; we used it to determine if followup was adequate.

  Approach: In general, we considered followup greater than or equal to 90 percent in the short term and 80 percent in the long term to be good.

• 6

  Statistical analysis: This factor included whether the investigators conducted the study in an appropriate manner and took the effect of multiple comparisons into account. This item also reviewed the study's use of multivariate statistical techniques and/or participant restriction or stratification to control for confounding.

  Approach: We rated this category on the basis of an average across all outcomes for each time point. Articles that reported appropriate statistical tests, point estimates, tests for homogeneity, stratification, and confidence intervals were rated as good. Articles that reported P-values alone were rated as fair, and articles that did not report statistical analysis were rated as poor.

Two article abstractors independently rated each article on each of the categories as indicated by the quality assessment form. We reconciled differences by consensus, giving each item equal weight. Specifically, articles that received good ratings on all categories would have been eligible to be rated as good studies overall. None of the three articles received a good rating. If an article received one or two fair or poor ratings, or the equivalent of a deficiency, it was rated as an overall fair-quality article. The original article from the Term Breech Trial, published in 2000, received a quality rating of fair.²⁰ Articles with three or more fair ratings or a poor randomization design or implementation with a fatal flaw were rated as a poor-quality article. The two follow-up articles from the Breech Trial, published in 2002¹⁸ and 2004⁸³ respectively, received a quality rating of poor.

We used the following criteria to rate the quality of nonrandomized observational studies:

• 1

  Study design: Given the difficulties of identifying planned cesarean delivery retrospectively, we assigned prospective studies a higher score.

  Approach: To receive a rating of fair for this component of study design, we required a study to be prospective.

• 2

  Study population: We sought documentation in the publication of the degree to which the study population was representative of women with uncomplicated spontaneous vaginal births in the study facilities or the broader population sampled.

  Approach: To receive a rating of good for this component of study design and conduct, we required a study to describe clearly (1) the base population from which cohort participants were sought, (2) the number of women in that base population (a denominator), and (3) the proportion of eligible women who were ultimately enrolled in the cohort.

  Studies with all three items were rated as good; studies lacking one item were classified as fair; and studies lacking more than one item were rated as poor.

• 3

  Comparability of subjects. For cohort studies, we sought five tiers of documentation showing that the study had (1) specific inclusion/exclusion criteria for all groups, (2) applied criteria equally to all groups, (3) comparable study groups at baseline with reference to variables not unique to mode of delivery, (4) study groups comparable to nonparticipants with regard to confounding factors, and (5) study groups comparable with regard to followup.

  In addition, for case-control studies, we sought documentation on whether the study had (1) explicit case definition, (2) case ascertainment not influenced by exposure status, and (3) controls similar to cases except that they did not have the condition of interest and did have an equal opportunity for exposure.

  Approach: We rated a cohort study as having good comparability of subjects if at least four of five elements were present. We rated studies as having fair comparability if two or three elements were present. Studies with fewer than two elements were rated as poor. We required case-control studies to have all three elements of the case-control rating to rate a good for the overall category. We rated case-control studies that were missing one element for the case-control rating as fair and those missing two or more elements rated as poor.

• 4

  Statistical analysis: We sought documentation on whether the study reported on the following aspects of statistical analysis: (1) appropriate statistical tests; (2) modeling and multivariate techniques or multiple comparisons; (3) power calculations and achieved sample size; (4) assessment of confounding by bivariate analysis, stratified analysis, or multivariable modeling; (5) reporting of adjusted estimates for main effects that took into account identified confounding or modifying factors (stratified or separate analyses were acceptable for simple constructs); and (6) presentation of adjusted results with a measure of statistical precision such as a confidence interval or P-value.

  Approach: We assigned a rating of good for the category of statistical analysis if studies provided at least five of the six elements above. We assigned a rating of fair if studies reported on three or four elements and a rating of poor if studies reported on fewer than three elements.

• 5

  Result and loss to followup: For all studies, we sought documentation on whether the study reported a measure of effect for outcomes and provided an appropriate measure of precision. In addition, for panel studies, we sought documentation of two follow-up measures: (1) analysis of how respondents differed from nonrespondents if loss exceeded 20 percent, and (2) if absolute loss to followup exceeded 25 percent.

  Approach: For studies with cross-sectional measures, we assigned a rating of fair if the study reported a measure of effect with an appropriate measure of precision. Studies without a measure of effect were rated poor. Panel studies needed to have an absolute loss to followup at or below 25 percent. If the differential loss to followup from panel studies exceeded 20 percent, the investigators needed to report on bias from followup to receive a good rating. We rated a study as poor for this component if it had more than 25 percent loss to followup or more than 20 percent loss without comparison for response bias.

For categories 1 and 5 above, studies could receive a maximum rating of fair. For categories 2, 3, and 4, studies could receive a maximum rating of good. We summarized the ratings across all five categories to assign an overall rating as follows:

• good, if the study received a fair on both categories with a maximum of fair rating and good on all three categories with a maximum rating of good;

• fair, if the study received three to five fair scores with fewer than two good scores; or

• poor, if the study received two or fewer fair or good scores.

Grading the strength of available evidence. Our scheme follows the criteria applied by West et al.⁸⁵ That system included three domains: quality of the research, quantity of studies (including number of studies and adequacy of the sample size), and consistency of findings. Two senior staff members assigned grades by consensus.

We graded the body of literature and present our findings in Chapter 4. The possible grades in our scheme are as follows:

• I

  Strong: The evidence is from studies of strong design; results are both clinically important and consistent with minor exceptions at most; results are free from serious doubts about generalizability, bias, or flaws in research design. Studies with negative results have sufficiently large samples to have adequate statistical power.

• II

  Moderate: The evidence is from studies of strong design, but some uncertainty remains because of inconsistencies or concern about generalizability, bias, research design flaws, or adequate sample size. Alternatively, the evidence is consistent but derives from studies of weaker design.

• III

  Weak: The evidence is from a limited number of studies of weaker design. Studies with strong design either have not been done or are inconclusive.

• IV

  No evidence: No published literature.

Literature Relevant to the Epidemiologic Questions

Overall, we identified 13 published studies reporting the incidence and trends of modes of cesarean section. Of the 13 articles, we gave 1 study a relevance rating of high,³³ 3 studies a rating of moderate,³⁴–³⁶ and 9 studies a rating of low.^{3, 56}–⁶³ We also found web-based sources for four regions: United Kingdom⁸⁶ and three states in Australia (Victoria,⁸⁷–⁸⁹ South Australia,⁹⁰ and New South Wales⁹¹). All of the Web-based sources were moderately relevant to CDMR. Two published studies were conducted in the United States,^{3, 56} 3 in the United Kingdom,^{33, 60, 61} 3 in Australia,^{34, 57, 58} 2 in the Republic of Ireland,^{35, 59} and 1 each in Norway,³⁶ Finland,⁶² and Denmark.⁶³ We also found web-based sources reporting rates of “elective” cesarean delivery for the United Kingdom⁸⁶ and three states in Australia (Victoria,⁸⁷–^{89, 92} South Australia,⁹⁰ and New South Wales⁹¹).

Four articles supplied trend data from the United States³ and Australia.^{34, 57, 58} Three of the four web-based sources provided trend data.^{86, 87, 91} With the exception of two published studies that obtained data from surveys usually sent to medical directors,^{59, 61} all other studies gathered data from administrative databases or materials (e.g., birth certificates). Except for one study,³³ all were conducted retrospectively.

Overall Estimates of Incidence and Trends

Rates of incidence and trends of cesarean delivery vary by country. In general, countries report rising trends of cesarean delivery, with recent incidence rates at 29 percent for the United States² and 23 percent for the United Kingdom.⁸⁶ In the following section, we present data on incidence and trends of cesarean delivery by country. We also present rates of primary prelabor cesarean, CDMR, cesarean for medical indications, and for malpresentation, when available; studies rarely provided sufficient information to answer this key question. Four studies report on primary prelabor cesarean,^{3, 34, 35, 56} and one study reports on CDMR.³³ Other studies use variable definitions of “elective” or “planned” cesarean delivery, denoted in quotes in the text. We present summary tables for each country, with each study listed in alphabetical order by last name of first author.

Country-Specific Incidence and Trend Data

United States. Incidence. In 2001, the rate of cesarean delivery was more than 25 percent.³ Studies from two sources suggest similar primary prelabor “elective” cesarean rates: 4.25 percent⁵⁶ and approximately 5 percent³ of all deliveries in 1995. The definition of elective primary cesarean delivery in both studies includes malpresentation, antepartum bleeding, herpes, severe hypertension, uterine scar, multiple gestation, macrosomia, unengaged head, soft tissue condition, other hypertension, preterm, fetal anomaly, and unspecified, resulting in a definition of “elective” cesarean delivery that has low relevance to CDMR (Table 6).

Trends. Meikle et al. reported a rise in the primary elective cesarean deliveries as a proportion of all deliveries from approximately 5 percent in 1994 to approximately 7 percent in 2001.³ The authors also reported a rise in elective primary cesarean deliveries as a proportion of all cesarean deliveries from 19.7 percent in 1994 to 28.3 percent in 2001.

United Kingdom. Incidence. One study from Scotland offers the only evidence of high relevance to CDMR that we identified.³³ From a prospective administrative dataset for 1994, the authors reported total cesarean deliveries (including multiple and unknown gestation) of 16 percent. The authors also reported a CDMR rate of 7.7 percent and an “elective” cesarean delivery rate for breech presentation of 19.1 percent of all singleton cesarean deliveries (Table 7). Two other studies from the United Kingdom did not define “elective” section, resulting in low relevance to CDMR. These studies used different sources. One study, using data from the Royal College of Obstetrics and Gynecology annual surveys, reported a total cesarean rate of 18.5 percent and an “elective” cesarean rate of 9.5 percent of all births from January 1997 to March 1998.⁶¹ The other study, using data from the National Health Service (NHS) episode statistics database, reported a total cesarean rate of 20.1 percent and an “elective” cesarean rate of 8.9 percent of all births for 2001-2002.⁶⁰

Trend. Web-based National Health Service data provide evidence that is moderately relevant to CDMR. They report trends for “emergency” and “elective” (defined as planned procedure before, or at the onset of labor) cesarean deliveries as a proportion of all births from 1990 to 2003-4.⁸⁶ During this period, the rate of all cesareans rose from 12 percent to 23 percent. Both elective and emergency cesareans contributed to this rise. Relative to all births, the rate of elective cesarean deliveries rose from 5.3 percent in 1990 to 9.6 percent in 2003, and the rate of emergency cesarean deliveries rose from 7.1 percent in 1990 to 13.1 percent in 2003.

Ireland. Incidence. Two studies using different sources reported data on Ireland for elective cesarean deliveries. One study, rated moderately relevant, provided incidence data from the National Maternity Hospital database annual report on the rate of all cesarean deliveries and “nonemergency prelabor” cesarean deliveries in 2000 as 15.1 percent. Over a 12-year period from 1989-2000, the authors reported a nonemergency prelabor cesarean delivery rate of 3 percent of all deliveries and 30 percent of all cesarean deliveries (Table 8).³⁵ Of these nonemergency prelabor cesarean deliveries, 24 percent were among primaparous women, 39 percent were primary cesarean deliveries among primiparous women, and 37 percent were among repeat cesarean deliveries. We calculated the total primary prelabor nonemergency cesarean rate to be 18.9 percent of cesarean deliveries during the 12-year period from 1989-2000. Farah et al. reported elective and emergency cesarean rates of 7.5 percent and 10.3 percent of all births, respectively, based on a retrospective survey of maternity unit directors.⁵⁹ The authors did not define “elective”; as a result, their study has low relevance to CDMR.

Australia. Incidence. The majority of studies, from both published and web-based sources, defined “elective” cesarean deliveries as a procedure planned before the onset of labor (Table 9).^{34, 87}–⁹² Two other articles either did not define “elective” cesarean or defined it as cesarean before and during labor, and hence did not exclude cesareans for fetal distress or other emergencies.^{57, 58}

The most recent figures from these studies indicate rates of all cesarean and elective cesarean delivery of 23.5 percent and 13 percent, respectively, in New South Wales in 2001;⁹¹ 27.4 percent and 14.1 percent, respectively, in Victoria in 2002;⁸⁷and 30 percent and 13.3 percent, respectively, in South Australia in 2003.⁹⁰

Trend. Trend data was reported on New South Wales,^{57, 58, 91} Victoria,⁸⁷–^{89, 92} and Western Australia.³⁴ Studies from the New South Wales database reported nearly constant rates of “elective” cesarean delivery or “cesarean before labor” over a period from 1990 to 1996⁵⁷-1997.⁵⁸ Web-based data for New South Wales from 1996 to 2001 showed a rise in the rate of all cesareans from 17.6 percent to 23.5.⁹¹ During this period, the rate of “elective” cesareans rose from 9.4 percent to 13.0 percent and the rate of “emergency” cesareans rose from 8.2 percent to 10.5 percent.⁹¹

Data from Victoria showed a rise in the rate of cesarean deliveries in a 10-year period from 1992 to 2002 from 17.7 percent to 27.4 percent in 2002.⁸⁷–^{89, 92} During this period, the rate of “elective” cesarean delivery rose from 9.7 percent of all deliveries to 14.1 percent and the rate of “emergency” cesarean deliveries rose from 8 percent to 13.3 percent.

One study from Western Australia reported on an earlier time period from 1980 to 1987. The study reported a cesarean delivery rate in 1980 of 11.2 percent; by 1987 it had risen to 16.9 percent.³⁴ The rate of “emergency” cesareans rose during this period from 5.8 percent to 8.3 percent; the elective cesarean rate increased from 5.4 percent to 8.7 percent.

The only study reporting rates of primary prelabor cesarean delivery used data from an administrative database from 1980 to 1987 in Western Australia. Using data provided in the publication, we calculated that primary prelabor cesarean deliveries rose from 4.1 percent of all deliveries in 1980 to 4.8 percent in 1987.³⁴ More recent data from another study indicate a rate of cesarean before labor of 2.5 percent for primiparous women in 1997 in New South Wales. The authors do not report primary prelabor cesarean rates for multiparous women.

Norway. Incidence. One study of moderate relevance to CDMR reported the rate of “elective” cesarean performed for feto-pelvic disproportion, breech, diabetes, hypertension, pre-eclampsia, twins, and low birthweight as 4.5 percent over a 10-year period (January 1986 to December 1995).³⁶ The rate of all cesarean deliveries during this period was 12.5 percent (Table 10).

Denmark. Incidence. The single Danish study reported a total cesarean rate of 11.9 percent and an “unplanned” cesarean rate of 7.7 percent of all deliveries in 1989 (Table 11).⁶³ The authors did not comment on planned cesareans, resulting in low relevance to CDMR; we infer that the remainder, that is, 4.2 percent, were planned cesareans.

Finland. Incidence. One Finnish study reported a total cesarean rate of 13.9 percent and “elective” cesarean rates of 7.1 percent, as a proportion of all births from July 1985 through June 1986 (Table 12).⁶² The authors did not define elective cesarean delivery resulting in low relevance to CDMR.

Maternal Outcomes for Primary Cesarean Deliveries

The following outcomes are relevant to both primary and subsequent cesarean deliveries. However, we draw upon evidence from studies focusing on primary cesarean deliveries to address the maternal outcomes listed below. As noted in Table 16, some of these studies include repeat cesarean deliveries. However, no study included in this review is limited to repeat cesareans. Outcomes particularly relevant to subsequent cesarean deliveries are addressed in the next section.

Mortality. Four studies reported on maternal mortality associated with mode of delivery (Table 17).^{20, 28, 64, 70} One is the randomized Breech Trial (relevance rating of T); we gave the initial report a quality rating of fair. This trial compared planned vaginal with planned cesarean for breech and analyzed results using intent-to-treat.²⁰ We rated another study as moderately relevant but of poor quality.²⁸ We rated the two remaining studies as having low relevance and did not rate quality.^{64, 70}

The Breech Trial initially reported one death in the planned vaginal group (n = 1,042, or 0.1%), and none in the planned cesarean group (n = 1,041).²⁰ The patient who died was described as “jaundiced before labor, developed disseminated intravascular coagulation after delivery, and died of hepatorenal failure at 44 hours postpartum” (p. 1380).²⁰

The moderately relevant (poor) study identified three maternal deaths that had been caused by underlying pathology.²⁸ The authors did not characterize these cases as either labored or unlabored cesarean deliveries, nor did they comment on indication for cesarean; thus, this study provides little information relevant to CDMR.

Of the two low relevance studies, one study reported some deaths but none associated with mode of delivery,⁷⁰ and the other reported no maternal deaths at all.⁶⁴

Infection. Twelve studies^{20, 28, 39, 45, 49, 51, 64, 67}–^{70, 84} included maternal infection as an outcome (Table 18).

The Breech Trial found no significant differences in the rates of wound infection or maternal systemic infection.²⁰ An earlier nonrandomized study (rated fair) compared a trial of planned vaginal with planned cesarean for breech and analyzed results using intent-to-treat. The investigators used a composite measure of maternal morbidity (febrile morbidity, endometritis, wound infection, urinary tract infection [UTI], and thrombophlebitis) and determined that it was significantly higher in the planned cesarean group.⁸⁴

Of five moderately relevant studies, two were of fair quality,^{39, 51} and three were of poor quality.^{28, 45, 49} Only one of these compared planned “intended” vaginal delivery with planned “elective” cesarean delivery.⁵¹ It did not give a detailed assessment of maternal infection beyond reporting a single case of sepsis among the 903 intended vaginal births and none among the 147 “elective” cesarean births.

Three studies compared outcomes only between various types of cesarean delivery; this restriction limited their utility for addressing the maternal infection issue in terms of planned CDMR vs. planned vaginal delivery. Of these, one study found no difference in UTI between planned “elective” and unplanned “emergency” cesarean deliveries but did find significantly lower rates of wound, intrauterine, and chest infections in the planned “elective” cesarean group.⁴⁵ The second reported no significant difference in the rate of UTI, wound infection, and endometritis between planned “elective” cesarean and unplanned “acute” and “secondary acute” cesarean deliveries.²⁸ The third set of investigators limited their analysis to wound infection and distinguished between major and minor infections. However, they did not provide a statistical comparison of wound infection rates between planned “elective” and unplanned “emergency” cesarean deliveries.⁴⁹

The most recent moderately relevant study compared various actual modes of delivery.³⁹ The risk of endometritis was significantly higher for both planned “without trial of labor” and unplanned “with trial of labor” primary cesarean deliveries than for spontaneous vaginal delivery. The risk of pneumonia was higher for both types of cesareans but significantly higher only for the planned “without trial of labor” primary cesarean group.

The remaining five studies were of low relevance because the authors either combined planned and unplanned cesareans in their cesarean comparison groups^{64, 70} or did not define “elective” cesarean delivery.⁶⁷–⁶⁹ Generally, these studies found that the risk of maternal infection was lower for planned “elective” cesarean than for unplanned or labored or “emergency” cesarean but lower for vaginal delivery than for planned “elective” cesarean.

Anesthetic complications. Of three studies reporting on anesthetic complications associated with mode of delivery, two were moderately relevant to CDMR^{50, 51} and one was of low relevance.⁶⁹

One moderately relevant study (fair quality) reported a 4 percent rate of problems with “peridural” anesthesia/postspinal headache in the “elective” cesarean group (6 of 147 women) and a 2 percent rate (18 of 903 women) with an intended vaginal birth;⁵¹ the authors did not provide statistical testing for this outcome. The other study (poor quality) reported no difference in anesthetic complications between planned cesarean and planned vaginal delivery.⁵⁰

The low relevance study obtained data from an administrative database using International Classification of Diseases, ninth edition (ICD-9) codes and reported rates of anesthetic complications in ascending order per 1,000 women: spontaneous vaginal deliveries, 90; assisted vaginal deliveries, 160; unplanned “nonelective” cesarean deliveries, 360; and planned “elective” cesarean deliveries, 390.⁶⁹

Hemorrhage/blood transfusion. Of the 11 articles (Table 19) that evaluated blood loss associated with route of delivery, 1 was from the Breech Trial,²⁰ 6 were moderately relevant,^{28, 38, 39, 45, 50, 51} and four were of low relevance.^{64, 67}–⁶⁹

As with other outcomes, definitions were not standardized. We encountered various measures of hemorrhage: blood loss >1,000 ml, blood loss >1,500 ml, blood transfusion, need for dilatation and curettage, and undefined postpartum hemorrhage. We chose to report these clinically relevant outcomes rather than anemia (change in hemoglobin or hematocrit level).

The Breech Trial (fair quality) included a heterogeneous group of women: women who were multiparous, had a history of a previous cesarean, and presented in labor.²⁰ The planned cesarean group had lower rates of postpartum bleeding (1.0%)—defined as estimated blood loss (EBL) >1,000ml, EBL >1,500ml, blood transfusion or need for dilatation and curettage—than the planned vaginal delivery group (1.3%); the difference was not statistically significant.

Five of the six moderately relevant studies limited comparisons to only planned “elective” vs. unplanned “emergency” cesarean delivery; they defined significant blood loss as blood loss >1000 ml,²⁸ transfusion,^{45, 50, 51} or both.³⁸ Of these five, two were of fair quality^{51 38}and three were of poor quality.^{28, 45, 50} The sixth study (fair quality) identified postpartum hemorrhage and transfusion through ICD-9 codes and reported these outcomes separately.³⁹

Of all six studies, five showed a lower risk of major blood loss with planned “elective” than with unplanned “emergency” cesarean delivery.^{28, 38, 39, 45, 51} The differences were statistically significant in two studies (both poor).^{28, 45}

The three moderately relevant studies that compared risk of blood transfusion differed in their comparison groups. One showed a nonsignificant higher absolute risk with planned cesarean (1.1 percent) than with unplanned cesarean (0.7 percent).³⁸ A second reported cases of blood transfusions in only the planned vaginal delivery group but provided no statistical testing.⁵¹ The third study found a higher risk of blood transfusion with both unplanned “with trial of labor” and planned “without trial of labor” primary cesareans than with spontaneous vaginal deliveries. The higher risk was statistically significant only in the unplanned “with trial of labor” group.³⁹

In the only study that did not report a lower rate of blood loss associated with elective cesarean birth,⁵⁰ the rates of blood transfusion were similar between elective cesarean and planned vaginal delivery: 1.2 percent vs. 1.1 percent, respectively.

Of the four studies of low relevance, three defined significant blood loss as requiring a blood transfusion;^{64, 67, 69} planned “elective” cesarean was associated with a lower rate of blood transfusion than that for planned vaginal delivery. Of these three studies, two compared planned “elective” cesarean with spontaneous vaginal, assisted vaginal, and unplanned “nonelective” cesarean or cesarean in labor.^{64, 69} In both studies, assisted vaginal delivery and nonelective or labored cesarean had the highest rates of blood transfusions. These results were statistically significant in only one study, which reported an absolute risk reduction for blood transfusion per 1,000 deliveries of 3.7 for nonelective cesarean, 1.2 for assisted vaginal, 1.1 for spontaneous vaginal, and 0.7 for elective cesarean delivery.⁶⁹ Other analyses comparing elective cesarean sections with uncomplicated vaginal deliveries suggested a significantly lower rate of postpartum hemorrhage among elective cesarean deliveries but no statistical difference in the rate of blood transfusion.⁶⁹

Hysterectomy. Three studies reported data on hysterectomy for postpartum hemorrhage.^{20, 38, 68} One was the Breech Trial article published in 2000 and rated of fair quality.²⁰ Another study that compared planned “elective” with unplanned “emergency” was moderately relevant and of fair quality;³⁸ the other was of low relevance.⁶⁸

The Breech Trial did not report any hysterectomies.²⁰ The moderately relevant study reported no significant differences in the rate of hysterectomy between elective cesarean (0.3%, 1 of 293 deliveries) and emergency cesarean (0.2%, 1 of 635 deliveries).³⁸ The low relevance study reported a single case of hysterectomy for hemorrhage in the elective cesarean delivery group.⁶⁸

Thromboembolism. Eight studies compared thromboembolism by route of delivery (Table 20). Of these studies, two were in the T relevance category (fair quality).^{20, 84} Two were moderately relevant studies, one of fair quality³⁹ and one of poor quality.²⁸ The remaining four studies were of low relevance.⁶⁷–⁷⁰

The studies universally lacked consistency in how they defined thromboembolism. Definitions varied from a composite outcome of maternal morbidity that included “thrombophlebitis” with other measures such as UTI, endometritis, and wound infection⁸⁴ to a single thromboembolic event measure that included obstetrical air embolism, amniotic fluid embolism, obstetrical blood clot embolism, other pulmonary embolism, cerebrovascular disorders, deep phlebothrombosis, and postpartum or unspecified venous thrombosis.⁶⁹ Two studies defined thromboembolic measures as deep venous thrombosis,^{39, 67} whereas another defined it as pulmonary embolism.⁶⁸ Yet another study assigned separate categories according to severity, defining “thrombosis” as a major morbidity outcome and “thrombophlebitis” as a minor morbidity outcome.²⁸

From the Breech Trial, Hannah et al. reported no cases of either deep vein thrombophlebitis or pulmonary embolism in either the planned vaginal delivery or planned cesarean delivery group.²⁰ Leiberman et al. had used the composite outcome defined above; contrary to the randomized breech trial, this group reported that this outcome was significantly higher in the planned cesarean groups than in the planned vaginal group.⁸⁴

One moderately relevant study with a vaginal delivery group found that the risk of deep vein thrombosis was higher among both planned “without trial of labor” and “unplanned “with trial of labor” primary cesarean deliveries; the risk was significant only for the unplanned “with trial of labor” cesarean group.³⁹ The other compared outcomes among planned “elective” and unplanned “acute” cesarean deliveries but did not contain a vaginal comparison group (a major limitation).²⁸ Thrombosis was part of a composite outcome of postoperative complications; that measure was significantly lower in the planned cesareans than in the unplanned cesareans (no statistical testing provided).

Of the four low relevance studies, three showed neither a significant difference nor a consistent direction of effect between planned “elective” cesarean and either vaginal delivery⁷⁰or planned vaginal delivery.^{67, 68} The remaining study, from the administrative data set noted above, reported the incidence of thromboembolic events in ascending order per 1,000 deliveries: 0.7, spontaneous vaginal deliveries; 1.1, assisted vaginal deliveries; 1.9, planned “elective” cesarean deliveries; and 4.5, unplanned “non-elective” cesarean deliveries. The rate of thromboembolic events was statistically higher in unplanned cesarean than in planned cesarean deliveries.⁶⁹

Surgical complications. The studies in this group are weighted toward surgical complications associated with cesarean deliveries. Our search parameters (reviewed by the Technical Expert Panel and the SOS Conference panel chair) were not designed to capture perineal and vaginal trauma associated with vaginal delivery. Therefore, we cannot provide a comprehensive assessment of the risks for perineal and vaginal trauma.

Of the 10 studies that compared surgical complications by mode of delivery, 1 was the Breech Trial,²⁰ 3 were moderately relevant,^{28, 38, 51} and 6 were of low relevance^{64, 67}–^{70, 75} (Table 21).

Similar to the practices seen for other maternal outcomes, we found some studies that defined surgical complications as a single composite measure of injury,^{64, 68, 69} and others that reported specific measures of complications such as bladder or bowel injury.^{20, 28, 38, 51, 67, 70, 75}

The Breech Trial (fair quality) found similar rates of genital tract injury (vertical uterine incision, serious extension to transverse uterine incision, cervical laceration extending to lower uterine segment, or vulvar/perineal hematoma requiring evacuation) among the planned vaginal group (0.6%) and the planned cesarean group (0.6%).²⁰ Neither group experienced genital tract fistulae, bowel obstructions, or injury to bladder, ureter, or bowel.

Two of the three moderately relevant studies were of fair quality^{38, 51} and one was of poor quality.²⁸ These three studies varied widely in their choice of comparison groups. The only moderately relevant study that compared planned “elective” cesarean with planned vaginal delivery “intended vaginal birth” focused primarily on perineal surgical injury.⁵¹ In this study, 33.4 percent of women in the planned vaginal delivery group experienced labial, vaginal, or first or second degree perineal lacerations; another 20.2 percent had an episiotomy; and 0.2 percent experienced a third or fourth degree laceration; as expected, no such complications occurred in the elective cesarean group. No abdominal surgical complications were reported for either the planned cesarean or the planned vaginal delivery group.

The two remaining moderately relevant studies were limited to comparisons of outcomes among cesarean deliveries.^{28, 38} One suggested a higher rate of surgical complications among women with unplanned (“emergency” or “labored”) cesarean delivery than among women with planned “elective” or “unlabored” cesarean;³⁸ the other did not.²⁸

Three of the six studies of low relevance to CDMR compared surgical complications between planned “elective” cesarean and planned vaginal delivery.^{64, 67, 68} All three reported slightly higher rates of surgical complications in the planned vaginal delivery group although these were not statistically significant.

Three other low relevance studies compared groups based on actual routes of delivery.^{69, 70, 75} One study using composite outcomes reported a higher rate of obstetrical (pelvic) trauma in spontaneous vaginal delivery (7.35 percent) and assisted vaginal delivery (7.05 percent) than in planned “elective” cesarean (1.09 percent) or unplanned “non-elective” cesarean delivery (0.57 percent). A subanalysis comparing planned cesarean to uncomplicated spontaneous vaginal delivery yielded similar findings. Two studies found that the rate of bladder injury was lower in planned “elective” cesarean than in unplanned “emergency”⁷⁰or “emergent” or “urgent” cesarean deliveries.⁷⁵

Breastfeeding. Two articles (rated poor) from the Breech Trial^{18, 83} provided evidence on initiation and duration of breastfeeding for women experiencing planned vaginal or planned cesarean delivery. The percentages of women initiating breastfeeding “within a few hours” were 77.6 percent for planned vaginal and 73.3 percent for planned cesarean; the difference bordered on statistical significance (P = 0.05).¹⁸ The percentages of women breastfeeding at 3 months and at 2 years were nearly identical.^{18, 83}

Postpartum pain. Four articles comprising three studies addressed postpartum pain.^{18, 43, 51, 83} Two articles reported later analyses for the Breech Trial (“T,” both rated poor).^{18, 83} The other two were moderately relevant, one of fair quality⁵¹ and one of poor quality.⁴³

The Breech Trial examined numerous pain outcomes at 3 months postpartum: any pain, location of pain, severity of pain, and use of analgesics during the past 24 hours and found no difference in the incidence of pain, severity of pain, or use of analgesics.¹⁸ As expected, at 3 months postpartum women in the planned vaginal delivery group were significantly more likely to report pain in the “bottom or genital area,” and women in the planned cesarean delivery group were more likely to report pain on the “outside of the abdomen” or “deep inside the abdomen.”¹⁸ These difference were no longer significant at 2 years postpartum.⁸³

Of the two moderately relevant studies, the fair quality study reported maternal pain using a visual analog scale (VAS) ranging from 1 (no pain) to 10 (most severe pain),⁵¹ that they administered at 3 days and at 4 months postpartum. At the 3-day postpartum evaluation, patients were also asked to rate retrospectively their pain at delivery. The authors presented results only in graphs, with no numerical outcomes for the VAS. They depicted a significantly higher median pain level during birth for the vaginal or assisted vaginal delivery groups than for the cesarean group. However, the authors commented that “peridural” anesthesia was offered to every woman but that only 11 percent chose it.⁵¹ At 4 months postpartum, they observed no significant difference in “momentary birth-related pain” among the groups. The investigators presented data for a group of women who underwent “cesarean on demand” for reasons including anxiety in nulliparous women, previous traumatic birth, coordination problems, and safety considerations. However, they did not report their data in a way that permitted us to abstract usable information, comment on laboring status in this group, or note specific analyses on this subset of patients.

The other moderate relevant (poor quality) study used a one-item Pain Intensity Scale within one week of delivery.⁴³ This study found no difference in mean pain intensity scores among unplanned cesareans, planned cesareans, and vaginal delivery.

Psychological outcomes: postpartum depression. Four articles comprising two studies dealt with postpartum depression associated with mode of delivery. The Breech Trial (rated “T”) contributed one article of fair quality²⁰ and two of poor quality;^{18, 83} the fourth study was of low relevance to CDMR.⁶⁶

The Breech Trial report (rated poor) defined depression as a score of more than 12 on the validated Edinburgh Postnatal Depression Scale.¹⁸ The low relevance study used the validated Center for Epidemiologic Studies Depression Scale. Neither study reported significant differences at any time point.^{18, 20, 83}

Psychological outcomes: other. Seven articles representing six studies reported on a variety of psychological outcomes other than depression; these studies used outcomes, measures, instruments, and time points for outcome measurement that have little in common. The two articles from the Breech Trial are rated “T” and are of poor quality,^{18, 83} two are moderately relevant,^{43, 51} and three are of low relevance.^{66, 76, 78} In general terms, women who experienced an unplanned cesarean birth or an instrumental vaginal delivery were more likely to experience adverse psychological outcomes than were women who underwent either a spontaneous vaginal or a planned cesarean birth.

The Breech Trial article reporting 3-month outcomes (rated poor) found that women in the planned cesarean delivery group were more likely than those in the planned vaginal birth group to indicate that they “liked being able to schedule their delivery” and “liked that childbirth experience was not very painful.”¹⁸ Women in the planned vaginal birth group were more likely than those in the planned cesarean group to indicate that they liked that the delivery was natural, liked actively participating in the birth, and liked that recovering from the childbirth experience was not difficult. These authors reported no differences between planned vaginal and planned cesarean births in women feeling reassured about their own health. In general, despite the differences in likes and dislikes, the planned vaginal and planned cesarean did not differ as to whether women would “participate in the trial if they had to do it all over again” (p. 1828).¹⁸ The follow-up study reporting outcomes at 2 years (rated poor) found no difference in the experience of being a mother, the relationship with husband or partner, or the relationship with husband or partner compared with that before the child was born.⁸³

Of the two moderately relevant studies, one fair-rated study⁵¹ used the Zerrsen test⁹³ for quantifying momentary personal feelings and a modified version of a birth experience questionnaire by Salmon and Drew.⁹⁴

These investigators reported no differences in momentary personal feelings before birth. At 3 days postpartum, women in the assisted vaginal delivery and emergency cesarean delivery groups reported strong negative feelings. These differences dissipated by 4 months postpartum. In contrast, women planning a cesarean delivery without medical indications had an expectation of a more pleasant birth than did women planning a vaginal delivery or a cesarean for medical indications. Of the 44 cesareans performed “on demand,” 20 (45 percent) were for women who had had a previous traumatic birth. At 3 days postpartum, the most positive birth experiences were reported by the group with planned cesarean without medical indications, followed in descending order by those with cesarean for medical indications, vaginal delivery, emergency cesarean, and assisted vaginal delivery. Results were similar at 4 months postpartum. The other moderately relevant (poor) study, using the Perception of Birth Scale⁹⁵ found no differences between the vaginal and planned cesarean delivery groups or the planned and unplanned cesarean groups.⁴³ However, women in the unplanned cesarean group had a more negative perception of the birth experience than did women in the vaginal group.

One low relevance study compared adaptive and ineffective responses during three time periods (1973-1980, 1981-1982, and 1989-1990). In the first and the last time periods, women in the unplanned cesarean group had a significantly lower percentage of adaptive responses and a higher percentage of ineffective responses than women who had a planned cesarean delivery. This finding was not statistically significant in the second time period.⁷⁶ A second study of low relevance appraised birth experience, neuroticism, and self-esteem.⁶⁶ None of these outcomes differed at 4 or 12 months among planned cesarean, unplanned cesarean, and vaginal delivery groups. A third study of low relevance to CDMR reported that women in the emergency cesarean and the assisted vaginal delivery groups had the most negative cognitions and emotions regarding the delivery overall compared with elective cesarean and normal vaginal delivery.⁷⁸

Maternal length of stay. Four studies reported on length of stay; one was the Breech Trial (fair quality)²⁰ and three were of moderate relevance (poor quality)^{28, 43, 50} (Table 22). The studies varied in comparison groups. Two investigated outcomes by intended route of delivery: planned vaginal vs. planned cesarean delivery,^{20, 50} and both reported a significantly higher median length of stay in the planned cesarean group. A third study reported length of stay separately for “planned” and “unplanned” cesarean deliveries and found significantly higher length of hospital stay among “unplanned” and “planned” cesarean compared with vaginal deliveries.⁴³ The fourth study limited outcomes to cesarean deliveries and found that the length of stay following a planned “elective” cesarean was shorter than among unplanned “acute” deliveries.²⁸ However, the authors provided no statistical test results. The results were consistent in demonstrating a longer hospital stay following planned or unplanned cesarean than following vaginal delivery.

Urinary incontinence. Nine articles comprising eight studies reported on urinary incontinence associated with mode of delivery (Table 23). Two articles were from the Breech Trial.^{18, 83} Two were moderately relevant,^{41, 44} and five were of low relevance.^{70, 73, 74, 79, 82}

The Breech Trial was designed primarily to focus on neonatal outcomes following planned vaginal vs. planned cesarean for breech. As such, it has significant limitations to outcomes related to pelvic floor disorders since the study included multiparous women, allowed randomization in labor, suffered from a high degree of crossover, was performed in 26 countries, used nonvalidated instruments in multiple languages, and more than 50 percent of participants required assistance in completing the questionnaires. The 2002 Breech Trial article (poor quality) suggested that planned cesarean delivery significantly reduced the risk of urinary incontinence compared with planned vaginal delivery at 3 months, with a relative risk of 0.62 (95% CI, 0.41–0.93).¹⁸ In the 2-year postpartum article (poor quality), the rates of urinary incontinence remained higher in the planned vaginal group than in the planned cesarean group, but the difference was no longer statistically significant.⁸³ The rates of urinary incontinence were about three times as high at 2 years in both groups as at 3 months postpartum. The authors suggested that a change in the reference period in the outcome measurement may explain this difference. At 3 months, women were asked about urinary incontinence in the past 7 days; by contrast, at 2 years, women were queried about the past 3 to 6 months.⁸³

The two moderately relevant studies (both fair) were prospective cohort studies; they investigated symptoms of stress urinary incontinence according to actual mode of delivery and adjusted for preexisting urinary incontinence. One study found that planned “elective” cesareans performed before labor and cesareans performed during the first stage of labor appeared to be significantly protective against urinary incontinence compared with spontaneous vaginal deliveries.⁴¹ Overall, the risk of postpartum urinary incontinence at 6 weeks was as follows: forceps, 35 percent; spontaneous vaginal delivery, 23 percent; cesarean in labor, 9 percent; and elective cesarean, 4 percent. In the other moderately relevant study, the prevalence of stress urinary incontinence 1 year postpartum was not significantly different among primiparous women who underwent a spontaneous vaginal delivery (10.3 percent) from the rate among women who had a cesarean delivery for obstructed labor (12.0 percent), but was significantly lower for women who underwent an elective cesarean (3.4 percent) (P = 0.02).⁴⁴

Three of the five low relevance studies compared symptoms of stress urinary incontinence by actual route of delivery. Two concluded that cesarean delivery had a protective effect relative to vaginal delivery.^{73, 82} The third showed lower rates of stress urinary incontinence for women who had a cesarean delivery compared with women who had a vaginal delivery; these results were statistically significant for primiparous women only.⁷⁹

The remaining two studies of low relevance to CDMR linked surgical administrative databases and birth registries to assess the association between route of delivery and surgery for stress urinary incontinence⁷⁴or hospitalization for either stress incontinence or “vaginal descensus.”⁷⁰ They yielded conflicting information about urinary incontinence issues.

Anorectal function. Of the seven articles comprising six studies (Table 24) that reported on anal incontinence associated with mode of delivery, two were from the Breech Trial.^{18, 83} Two others were moderately relevant,^{42, 46} and three were of low relevance.⁷⁰–⁷²

Six articles assessed symptoms of anal incontinence.^{18, 42, 46, 71, 72, 83} The seventh, a population-based study, linked an administrative database and birth registry of primiparous women who delivered singleton breech infants at term to assess the association between route of delivery and anal sphincter rupture over a period of up to 23 years.⁷⁰

Of the six studies reporting on symptoms, all but one included flatal incontinence in addition to involuntary loss of solid or liquid stool in their definition of anal incontinence. The remaining study limited its definition of anal incontinence to frank incontinence and fecal urgency.⁷¹

Three studies assessed women for preexisting anal incontinence.^{42, 46, 71} No study used a validated instrument. The time period for the assessment of anal incontinence ranged from 3 months¹⁸ to 2 years.⁸³

The two Breech Trial reports (rated poor)^{18, 83} used different measures at two time points; questions related to fecal incontinence were added after some participants had already completed the study. At 3 months, the authors queried participants regarding whether they had experienced fecal incontinence in the past 7 days.¹⁸ At 2 years, however, the participants were asked about fecal incontinence over the previous 3 to 6 months.⁸³ Neither article reported a significant difference in rates of fecal incontinence between planned vaginal and planned cesarean.

The two moderately relevant studies were of fair quality.^{42, 46} One reported new onset symptoms of anal incontinence in 3 of 80 women (3.8 percent) in the planned “elective” cesarean group, 6 of 104 women (5.8 percent) in the unplanned “emergency” cesarean group, and 8 of 100 (8 percent) in the vaginal delivery group.⁴⁶ This progression suggested an increasing risk of fecal incontinence with emergency cesarean and vaginal delivery compared with elective cesarean. However, the authors limited their statistical comparison to overall cesarean (elective and emergency) against vaginal delivery (P = 0.427). They also noted a higher risk of severe fecal incontinence after elective cesarean, 2 of 80 women (2.5 percent), than after emergency cesarean, 1 of 104 women (0.96 percent); the latter rate was similar to that for vaginal delivery, 1 of 100 women (1 percent). The authors suggested that elective cesarean is not always protective and that symptoms of fecal incontinence associated with elective delivery can be severe. However, these results need to be interpreted with caution given the low incidence rates overall.

The other moderately relevant study reported a significantly lower rate of flatal incontinence at 6 months in the planned “elective” cesarean group (0 percent) than in the unplanned “in labor” cesareans (21 percent).⁴² The authors of the study noted that the risk of flatal incontinence was higher with forceps-assisted delivery than with spontaneous vaginal delivery and that the risk of both flatal and fecal incontinence was higher in both groups than in all cesarean deliveries.

Of the three low relevance studies, one reported that primiparous women had no cases of anal incontinence in the elective cesarean group (0 of 13 women).⁷¹ For the other modes of delivery, the rates of anal incontinence in descending order were as follows: vacuum delivery, 3 of 14 women (21.4 percent); emergency cesarean, 5 of 59 women (8.5 percent); forceps delivery, 5 of 86 women (5.8 percent); and spontaneous vaginal delivery, 5 of 189 women (2.6 percent). Similarly, among multiparous women, no case of anal incontinence occurred in the elective cesarean group (0 of 48 women). The overall rates of fecal incontinence in descending order were as follows: vacuum delivery, 1 of 4 women (25.0 percent); forceps delivery, 3 of 24 women (12.5 percent); spontaneous vaginal delivery, 13 of 379 women (3.4 percent); and emergency cesarean, 1 of 54 women (1.9 percent). The authors reported no statistically significant difference among groups and attribute this to small numbers. The authors performed logistical regression modeling and found that both vacuum and forceps were statistically associated with anal incontinence, P = 0.002 and P = 0.027, respectively.

Another low relevance study was a prospective questionnaire study comparing symptoms and actual route of delivery.⁷² Cesareans overall had a lower risk of fecal incontinence than did spontaneous vaginal deliveries (OR = 0.58; 95% CI, 0.35–0.97). This study also found a higher risk of fecal incontinence associated with forceps deliveries than with spontaneous vaginal delivery (OR = 1.94; 95% CI, 1.30–2.89).

The other low relevance study from the administrative database reported no case of anal sphincter rupture in either the elective (n = 7,503) or emergency cesarean groups (n = 5,575).⁷⁰ They did report, however, 41 of 2,363 cases (1.7 percent) in the vaginal delivery group. No group had any hospitalizations for either anal incontinence or fistula.

Pelvic organ prolapse. One low relevance study that examined pelvic organ prolapse associated with various actual modes of delivery, using an administrative data set compared hospitalizations for either vaginal descensus or urinary incontinence between 5 and 18 years after delivery.⁷⁰ The publication appears to have a typographical error. The rate of hospitalization for either prolapse or urinary incontinence was reported as 42 of 7,503 (0.6%) for elective cesarean, 13 of 2,363 (0.55%) for vaginal delivery, and 80 of 5,575 (1.4%) for emergency cesareans. However, the actual manuscript reports the rate of hospitalization in the emergency cesarean group as “80/5575 (0.5%).” The authors report that difference in hospitalization for either pelvic organ prolapse or urinary incontinence did not differ statistically, but they state that they are unsure how to interpret these results accurately. The potential error seems limited to the emergency cesarean group.

Sexual function. Two articles comprising the Breech Trial,^{18, 83} received a relevance rating of T, were of poor quality, and included sexual function outcomes. Sexual function was measured differently at three months and two years after delivery. At three months, measures included sex since birth and pain during sex on most recent occasion. At two years, measures included aparuenia, dysparuenia, the presence and extent of sexual problems and happiness with sexual relations. No statistically significant differences were noted at either time point for any measure. However, we note that none of the measures was validated and the instruments were administered in multiple languages and with the assistance of translators.