The Term Breech Trial Ten Years On: Primum Non Nocere?

Abstract

Abstract:  In 2000, the Term Breech Trial was published, and its authors recommended cesarean section as the safest mode of delivery for breech-presenting babies. Criticisms of the trial were raised at the time, which the authors dismissed. Since then, maternal deaths have been recorded among women undergoing cesarean sections for breech presentations. Accordingly, those initial criticisms deserve to be revisited. (BIRTH 39:1 March 2012)

It is now over 10 years since publication in The Lancet of the study known as the Term Breech Trial by Hannah et al, in which a planned vaginal breech delivery was compared with a planned cesarean section (1). The authors concluded that cesarean section is the safest mode of delivery for the baby. This publication resulted in an immediate swing to the performance of elective cesarean section for breech presentations in most countries, although some American obstetrics units, such as in California, already had an incidence of over 95 percent (2). The trial was stopped early on the advice of an independent monitoring panel, after it seemed that there was increased mortality and morbidity among the vaginal breech babies.

In September 2004, the author’s publication of a 2-year follow-up of children from the study reported no difference between the two groups in terms of morbidity or mortality. They stated that “This is in contrast to our earlier report” (3, p 869). Nonetheless, they still argued that cesarean section was the safest method for breech delivery (4). It is very strange that a research group can report no long-term difference between two management protocols, and yet still insist that one treatment regimen is superior to the other. It is as though the authors could not accept that their original recommendation—which they had defended in emotional language: “the lives and health of babies are at risk” (5, p 219)—was now unsubstantiated.

The downside of recommending elective cesarean sections for breech presentations is that women will now incur the attendant risks of the operation, both short- and long-term. Since their study was published, five maternal deaths have been reported in the literature of women who underwent cesarean section with a breech presentation (6,7). The deaths that have been reported to date are almost certainly an underrepresentation of the real incidence of maternal mortality, because achieving publication of an occasional maternal death after a breech cesarean section is difficult.

The main long-term complication is the likelihood of repeat cesarean sections, with the risks of ruptured scars, placental invasion of the uterus, and hysterectomy (8,9). In view of the emergence of maternal deaths, however infrequent, clinicians should review the original report in depth.

Background of the Trial

In 1994, a Consensus Conference in Toronto developed criteria for the study (10) and a protocol was published (11). Women were eligible for the trial if they had a singleton live fetus in a frank or complete breech presentation at 37 weeks’ gestation or more, and were randomized to either planned cesarean section or planned vaginal breech delivery. Women were excluded if there was evidence of fetopelvic disproportion, if the fetus was judged to weigh 4 kg or more, if the fetal head was hyperextended, if a fetal anomaly or condition was present that might cause a mechanical problem at delivery, if a contraindication to either labor or delivery (e.g., placenta previa) existed, or if the fetus had a known lethal congenital anomaly.

Vaginal breech births were to be performed by an “experienced” clinician—someone who considered himself or herself to be skilled and experienced at vaginal breech delivery, with confirmation by the clinician’s head of department. Many of the clinical guidelines would appear to be at odds with standard obstetric practice. In the published results, 3½ hours was permitted for the second stage of labor (1, p 1376). In addition, 54 women in the study had undergone a previous cesarean section (12, p 227), which many clinicians would consider made them not “good candidates” for a vaginal breech delivery.

One hundred and twenty-one hospitals from 26 countries participated in the trial. The countries were categorized as having either a “high” (n = 10) or “low” (n = 16) perinatal mortality rate. Maternity units participating in the study were classified as having either a “high” standard of care (35.2%); or a “usual” standard of care (64.8%). Criteria that defined a unit with a “usual” standard of care included having no one who could provide oxygen by mask within 10 minutes of delivery or endotracheal intubation to an asphyxiated baby within 30 minutes. (Many clinicians would use the phrase “substandard care” if their maternity unit could not administer oxygen by mask within 10 minutes or endotracheal intubation within half an hour.)

Participating Third World Countries

Enrolling hospitals from Third World countries may well have helped boost the numbers needed for the study. However, any study is duty-bound to compare like with like. It is not disrespectful to health workers in resource-challenged countries to state that their doctors, nurses, and midwives work in circumstances that differ significantly from units in most Western centers. Without any disrespect to health care workers in the Third World, are obstetric and perinatal outcomes in, for example, Zimbabwe, comparable with those in Switzerland? The authors later revealed that ultrasound facilities were not available in “many of the participating centers” (5, p 218). Ultrasound is a basic tool of a modern obstetric care. If many units were lacking ultrasound scan facilities, how many were also lacking cardiotocograph and sonicaid facilities?

Analysis

The results of the randomized controlled trial were analyzed by “intention to treat.” The randomized controlled trial together with intention to treat can be portrayed by enthusiasts as the gold standard to end all gold standards, capable of analyzing any uncertain issue: “If you find that a study was not randomized, we’d suggest that you stop reading it and go on to the next article” (13, p 118). However, no statistical method of analysis is without its limitations. The limitations of randomized controlled trials have been described by many authors (14–18). Black observed that fervent randomized controlled trial believers are likely to regard disbelievers as “methodologically incompetent” (15). The downside of a randomized controlled trial is that, if poorly monitored, it can become a fishing expedition, in which every conceivable event may be included, however irrelevant, and that disconnected adverse outcomes can distort the results.

Intention-to-treat analysis is a method of seeking information about the potential effects of the treatment policy, rather than that of the actual treatment itself. Bandolier noted that the phrase Intention to Treat has different meanings for different authors (19). Hollis and Campbell argued that one of the key issues in intention-to-treat analysis is the handling of trial violations, be they violations of patient selection protocol or violation of the treatment protocol (20). They believed that researchers should justify in advance any selection criteria, which, if violated, would merit exclusion from intention-to-treat analysis, and that handling of protocol violations should be explicitly described in the reporting of results. However, the authors of the Term Breech Trial provided no comment about their policy of handling violations, and all their violations were simply included in their final analysis.

Results: Adverse Outcomes

Entry and outcome data were ultimately received for 2,083 women. Of the cesarean section group (n = 1,041), 941 (90.4%) were delivered by cesarean section. The remainder had a vaginal delivery because the baby’s birth was imminent (n = 59), because of the patient’s request (n = 29), the baby was in a cephalic position (n = 12), or for reasons unknown (n = 5).

Of the planned vaginal delivery group (n = 1,042), 591 delivered vaginally (56.7%). The reasons why 451 women had cesarean sections were as follows: fetopelvic disproportion or failure to progress in labor (n = 226); fetal heart rate abnormalities (n = 129); footling breech presentation (n = 69); patient’s request (n = 61); obstetric or medical complications (n = 45); cord prolapse (n = 12); and reasons unknown (n = 8).

Perinatal Mortality

After removing 5 infants with congenital defects, the study reported that 13 of the planned vaginal birth group babies died, but only 3 from the planned cesarean section group. The clinical details of the 16 perinatal deaths, both in the published article (1) and in its associated website, were sparse. The cause of death was not provided. Nonetheless, examination of those details that were made available indicate that many of the perinatal deaths appeared to have no connection with the mode of delivery or labor, and that most of the “serious” neonatal morbidity markers reflected transient events, or were incorrect diagnoses. For example, these infants were among the 16 perinatal deaths:

• Case #2.  Macerated stillbirth, 1,125 g; “probably” died before recruitment.

• Case #6.  Discharged home well, died in his or her sleep (sudden infant death syndrome possible).

• Case #7.  Baby had a small head, low-set ears, deep-set eyes (congenital defect possible).

• Case #9.  Discharged home well; died after severe vomiting and diarrhea (gastroenteritis possible).

• Case #15.  Intrauterine death; “probably” died before recruitment.

It is remarkable that mothers could be recruited into the study, without it being established whether the baby was actually alive or dead. To attribute the cause of death of babies, already dead before recruitment, to their subsequent delivery is simply untenable. In the cases listed before, case #2 further violated the protocol in that it was a twin, not a singleton pregnancy. Case #15 was not even a breech presentation, but a cephalic presentation; yet by the randomized controlled trial rules, it was still called a vaginal breech death.

Apart from the five cases described before, two babies who weighed 2,500 and 2,700 g, respectively, died because of “respiratory problems.” No further information was provided. Was it hyaline membrane disease? A respiratory tract infection? Without further clinical details it is impossible to connect their deaths to either their mode of delivery or the management of the labor.

Serious Neonatal Morbidity

It was reported that 197 markers of neonatal morbidity were among the planned vaginal group, but only 61 among the planned cesarean section group. The authors concluded that “we have shown that a policy of planned cesarean section is substantially better for the singleton fetus in the breech presentation at term” (1, p 1382). The original trial report listed what it described as “neonatal morbidity” (1, table 6, p 1380). Many of these instances appear to have been very transient events. For example, 18 babies in the vaginal breech group were described as suffering from “hypotonia”; in seven cases, the “hypotonia” disappeared within 2 hours.

At the 2-year follow-up, the authors conceded that “most children with serious neonatal morbidity survive and develop normally” (3, p 870), which raises the question of whether the neonatal morbidity markers in the trial were incorrectly or prematurely applied.

Equally troubling are diagnoses of “morbidity” that turn out to be incorrect. In the original 2000 report, a baby was described as having a spinal cord injury during a planned cesarean section (1, table 6, p 1380). However, 4 years later, it was revealed that this baby merely experienced “hypotonia” for 24 hours (4, table 3, p 1070). Normally, several months occur before publication for the authors to make certain that their facts are correct. Such a belated revelation of such an elementary mistake makes it difficult to have confidence in the study’s results.

Protocol Violations

Some selection criteria in the trial protocol that were violated included the following: 1) live baby, 2) singleton baby, 3) experienced clinician, and 4) nonfootling breech.

• 1
   Live baby. As described earlier, among the vaginal breech deaths, it appears that two babies (#2 and #15) were “probably dead” before they were enrolled into the study. The Methods section of the trial report states: “Women were eligible for the trial if they had a singleton live fetus” (1, p 1375). Yet, these two prestudy deaths were still included as “vaginal breech delivery deaths.”
• 2
   Singleton baby. As mentioned earlier, case #2 was a twin pregnancy.
• 3
   Presence of an “experienced” clinician. The protocol (10) stressed three times that a vaginal breech delivery should be performed by a clinician experienced in vaginal breech delivery. In the published article, it reads: “The study was done in centres that could assure women having a vaginal breech delivery that an experienced clinician would be present at the birth” (1, p 1375). Yet, the authors reported (1, table 3, p 1378) that 21 vaginal deliveries were performed by someone other than an “experienced” clinician. These cases were still included in the final analysis.
• 4
   Nonfootling breeches. The protocol stated that to be included in the study, the presentation of the fetus must be either frank or complete breech. Yet, 12 babies with either footling or “uncertain” breech presentations were delivered vaginally and included in the final results.

These violations are presented as of no great concern but, rather, as some sort of a triumph of quality control. Hannah et al reported that “only” 58 of the 646 women who had vaginal deliveries violated the protocol for a vaginal breech delivery, which was described as “excellent” compliance (1, p 1379). Lumley uncritically commented that compliance in the study was “exceptionally good” (21).

Quality Control Issues

In 1994, the editors of the New England Journal of Medicine commented on the “difficulty of maintaining surveillance … in large clinical trials … over widely dispersed, heterogeneous institutions” (22, p 1449). Supervising quality control of 121 maternity units in 26 countries was always going to be challenging. It may be inevitable that in over 2,000 deliveries some clinical events will “push the envelope” of acceptable care. Nonetheless, some of the reported clinical incidents are very disturbing. For example, a baby who was delivered by an “experienced” obstetrician in a planned prelabor cesarean section, sustained a basal skull fracture (4, table 3, p 1070). Another baby booked for a planned cesarean section by an obstetrician, had a prolonged labor (i.e., >18 hr), yet was still delivered vaginally, as a footling breech (4, table 3, p 1070).

Bizarre Prediction

The Term Breech Trial authors claimed that their study suggests that as few as seven extra cesarean sections are needed to prevent a dead or compromised baby in low perinatal mortality rate countries (1, p 1382). In the same issue of The Lancet, Lumley unfortunately repeated this claim (19). Subsequent Dutch studies concluded that it would take either 175 or 320 cesarean sections (23,24) to save one baby from death or severe compromise. A Swedish study suggested it would take up to 400 cesarean sections (25). The trial team has never modified its original prediction.

Premature Cessation of the Term Breech Trial?

The original plan was to recruit 2,800 women for the trial. An analysis was performed after results from 1,600 women were available. After this review the monitoring committee advised that the trial be stopped, as they thought that the difference in the adverse primary outcome between the two groups was significant. Recruitment was stopped the next day. However, additional data became available for 483 women, so that the total number of women for analysis was 2,083. The early termination of the trial was portrayed by the authors as though this act in itself was somehow compelling proof that vaginal breech deliveries had been proved to be unsafe. In view of the 2-year follow-up results, it is now legitimate to ask some questions: What information was provided to the monitoring committee? Were they presented only with the statistics? Were they informed of the multiple protocol violations, and that many of the perinatal deaths in the vaginal breech group were not relevant or connected in any way to the mode of delivery or labor? Were they informed that two babies were almost certainly dead before recruitment? Were they aware that much of the so-called serious neonatal morbidity (such as the “spinal cord injury” described earlier) was either a transient event or a misdiagnosis?

Other Studies on Breech Births

Since 2000, several publications from around the world have reported their own hospital’s or their nation’s experience with breech presentations, some advocating cesarean section and others advocating carefully selected vaginal breech delivery. Those that advocate planned cesarean sections include Rietberg et al (23, The Netherlands); Krebs et al (26, Denmark); Gilbert et al (2, California); and Herbst (25, Sweden). Those that support carefully selected and monitored vaginal breech birth include Alarab et al (27, Ireland); Doyle et al (28, Texas); Kumari et al (29, United Arab Emirates); Goffinet et al (30, France and Belgium); Krupitz et al (31, Austria); Uotila et al (32, Finland); and Haheim (33, Norway).

Goffinet et al’s review of vaginal breech delivery in France and Belgium, called the PREMODA (Presentation and Mode of Delivery) study, analyzed over 8,000 deliveries, almost four times the number in the Term Breech Trial, and did not find any significant excess risk associated with planned vaginal breech deliveries (30). Rietberg et al’s results from the Netherlands indicated decreased perinatal mortality and morbidity among breech babies having cesarean sections, which the authors attributed to the rapid increase in cesarean section rates since the Term Breech Trial (23). Other workers have not experienced the same result. In 1982, Green et al in Montreal reported that although they had increased the cesarean section rate from 22 to 94 percent between 1963 and 1979 for breech babies, this change had “not reduced unfavourable outcome significantly” (34, p 647).

The Term Breech Trial provided minimal case details. It would be useful if writers who advocate planned cesarean sections gave more clinical details of perinatal deaths within their studies.

The Healthy Term Breech Baby?

Some authors may believe that 100 breech-presenting babies at term are equivalent in health to 100 babies with a cephalic presentation. Brenner et al in 1974 appear to be the first to document the increased risk of congenital defects among breech deliveries compared with nonbreech deliveries—6.3 percent versus 2.4 percent (35). Another curious aspect of breech-presenting babies is their lower birthweight compared with nonbreech babies. A Swedish study in 1986 reviewed 228 consecutive singleton breech pregnancies at 33 weeks’ gestation, of which 132 turned spontaneously, and presented at delivery as cephalic, whereas the rest remained breech (36). At delivery, a significant difference in birthweight was reported between the two groups—the mean birthweight for the breech-presenting babies was 3,190 g, whereas those who presented cephalically had a mean birthweight of 3,595 g. The difference could not be explained. This reduced intrauterine growth may have reflected some underlying ongoing pathology, which was not understood, but would not be eliminated by the mode of delivery.

Response to Critics

The flaws in the trial listed previously have already been identified by other writers (37,38). In response to criticism, the trial authors have not been shy in dismissing their critics as being clinicians with closed minds, clinging to their prior beliefs. Replying to Keirse’s adverse comments (39), they wrote: “We conclude that Keirse’s disappointment in the trial has prevented him from being objective”; and “It is human to discount study results that run counter to one’s firmly held beliefs” (5, p 218).

Marek Glezerman, one of the trial collaborators, became so disillusioned with the study and its conclusion that he called for its complete withdrawal (40). In response, the trial team, without responding to any one of his detailed criticisms, merely commented that obstetricians who believe in vaginal breech deliveries have been the “most disappointed” (41).

It is possible that it is the Term Breech Trial team that has to cope with disappointment. After years of planning and organization, the team prematurely shut down the trial on what it strongly believed was compelling evidence, which was later found to be unsubstantiated. Now reports are emerging of maternal deaths that are associated with the policy that they have promoted.

Accountability

However rare an event, a maternal death is an overwhelming catastrophe. When it happens after a breech cesarean section, who exactly is accountable? The obstetrician who performs the operation? The hospital that prohibits planned vaginal breech deliveries? Those who actively promote such a practice? Or is it just collateral damage, which the relatives of the dead woman have to accept, and no one individual can be held responsible for the outcome?

The Inevitable Persistence of Vaginal Breech Deliveries

Vaginal breech deliveries will never completely disappear for three reasons: First, breech presentations can still be undiagnosed prior to labor. Second, some women who have chosen to have a planned cesarean section arrive in advanced labor, too late to have the cesarean section, which in the trial occurred in 5.7 percent of the planned cesarean section group. Third, some women, having done their own research, have made a considered decision not to have a cesarean section. One little-appreciated scenario is the woman who originally undergoes an elective cesarean section in an early pregnancy for a breech presentation. She then has repeat cesarean sections for her future deliveries. After several cesarean sections, she eventually runs the risk of a life-threatening placenta percreta—a scenario that is already a reality.

The Future

Unfortunately, the Term Breech Trial, with its multiple violations, its dubious quality control, and incorporation of perinatal deaths unconnected to either the labor or the mode of delivery cannot be regarded as a satisfactory or definitive answer to the question of the safest management for delivery of a breech-presenting baby at term. As far as the incidence of maternal short- and long-term complications are concerned, it may be necessary to establish a database to monitor such outcomes. This process could be undertaken by the international bodies that survey infrequent pregnancy outcomes, such as the International Network of Obstetric Survey Systems, the UK Obstetric Surveillance System, or the Australian Maternity Outcomes Surveillance System.

Realistically, the trial’s publications, despite the deficiencies described here, have accelerated the trend to cesarean section for breech deliveries almost to the point of no return.

In 2007, Chinnock and Robson reported on the future practice intentions of Australian obstetric trainees in their final year of training (42). Only 2 of 19 planned to offer vaginal breech deliveries to their patients, which means that a new generation of obstetric accoucheurs is emerging who are increasingly unskilled in vaginal breech deliveries, and leaving those remaining vaginal breech babies born unexpectedly to be more at risk than ever.

The Australian Institute of Health and Welfare publishes annual figures on confinements in Australia (43,44). In 1991, a total of 10,432 breech presentations at delivery were reported. Of these pregnancies, 2,884 women went on to have vaginal breech births (27.6%). In 2008 there were 11,816 breech presentations at delivery; of these, only 1,642 women went on to have vaginal breech deliveries (13.9%). Even with this increased cesarean section approach, it still means that, each week, somewhere in Australia, approximately 30 breech babies are being delivered vaginally. Who will be skilled to handle them?

In the future, there may well be two types of obstetricians—those who are competent to manage vaginal breech deliveries safely and those who are not. To avoid such a situation, it is surely the responsibility of the national obstetrical training colleges to ensure that all trainee obstetricians have the capacity to handle an unexpected vaginal breech delivery.

In an editorial entitled Teaching Infrequently Used Skills, Queenan argued that performing an infrequent vaginal breech delivery may have to be seen in the same light as the compulsory teaching of cardiopulmonary resuscitation, even though it is rarely needed (45). To achieve this, he called for the use of training films, hands-on sessions, and mannequins. Some countries are already using these approaches. Obstetricians in the Netherlands have run a “Dutch Course of Breech Delivery” since 2003. In addition, protocols in units with established vaginal breech programs—such as the PREMODA study—should be carefully analyzed to see what lessons can be learned. Otherwise, in the post-Term Breech Trial era, for women who present in advanced labor with a breech presentation, it may well be “pot luck” that they will have a doctor or a midwife who hopefully knows what he or she is doing.

Acknowledgments

The author thanks Andrew Bisits and David Ellwood for their reviews of an early draft of this manuscript.