The ‘Term Breech Trial’ and its aftermath is a prime example of how evidence itself can be put on trial.
Hannah and colleagues’ so-called ‘term breech trial’ (TBT) was published over ten years ago now1, and we still practise in its shadow. Many experienced clinicians urged caution in accepting the study’s initial outcomes; data from the two-year follow-up of the babies was reassuring, revealing no significant differences in the major outcome measures with respect to death or neuro-developmental delay.2 Time has told a different story, however, and the trend to deliver term singletons in the breech presentation by caesarean section has continued unabated.3
Why did we urge caution in accepting the term breech trial findings?
Some clinicians and institutions who were originally approached to participate in the term breech trial had expressed concern that the trial protocol was asking participants to undertake practices that were at the least unusual and, at best, not at all part of the routine approach to breech delivery in Australia and/or New Zealand at that time. I would summarise these issues as follows:
- Although the breech delivery was to be ‘planned’, consent to enter the trial could be obtained while the woman was actually in labour, leaving little or no time for fetal assessment or maternal counselling.
- Cardiotocographic (CTG) monitoring in labour was optional, even though the delivery was deemed ‘high risk’.
- The vaginal delivery (VD) rate for the breech delivery arm was to be more than 50 per cent even though that rate was achieved in only a very few selected specialist breech units in the developed world4, and to achieve this oxytocin augmentation was to be utilised for delay in labour.
- Included were women who had been delivered by caesarean section as well as babies with a weight of less than 2.5kg, even though this weight is associated with growth restriction at term.
- Randomisation was to take place at or after 37 weeks, with the consequence that consequential numbers of fetuses would exceed the recommended upper range of weight of 4kg by the time labour commenced.
- Fetal weight and attitude of the head (the degree of flexion), although considered important as entry criteria, could be assessed clinically if no ultrasound was available.
- The degree of experience and training of the accoucheur appeared to be highly variable, including not only doctors with 20 years’ experience, but also individuals who deemed themselves experienced, and who had their head of department attest to this, regardless of whether or not they had been observed as being competent.
- The enrolment of women was to be ad hoc rather than consecutive. This left the potential for units to aim to deliver those women they thought highly likely to have a successful vaginal delivery and to randomise only the group in whom the prospect of a vaginal delivery was uncertain.
- Units would not contribute equal numbers of caesarean and vaginal births (there was no blocking of the units by delivery). This had the consequence that one unit with an obstetrician highly competent in breech delivery could be made to deliver all the babies of women enrolled in the trial by caesarean section.
- Units unable to provide emergency caesarean delivery within the hour were included. This also applied to units unable to provide oxygen to a baby for up to ten minutes after birth, and/or intubate a baby for up to 30 minutes.
It seemed apparent to many that the trial protocol was not necessarily optimal to answer the question as to whether: a mother with a singleton breech presentation at term in Australia or New Zealand, who has a complete pre-labour fetal and maternal assessment; has adequate and experienced counselling; can have continuous CTG monitoring in labour in a unit that can provide timely caesarean delivery and has neonatal resuscitation facilities; and can be attended by a well-trained, experienced and enthusiastic obstetrician, is better delivered vaginally or by planned caesarean section.
The investigators tried to account for the differences in facilities and standards by separating those centres with a perinatal mortality rate (PNMR) of less than 20/1000 births from those with a PNMR of more than 20/1000. A PNMR of less than 10/1000 is now common in units in the developed world and such units still differ markedly in their ability to mount a rapid caesarean and neonatal resuscitation response when compared to those with a PNMR of more than 10/1000.5-7 Ultimately, this subanalysis by dividing the units participating in the term breech trial was probably meaningless.
Undertaking large clinical trials can be a little like politics – full of compromises and confessions. Everybody understands this. Given the trial as published, and the subsequent published data flowing from it, are we still justified in saying that the term breech trial indicates the superiority of planned caesarean section over trying for a vaginal delivery? Numerous publications have critiqued the term breech trial3-7, and space constraint allows only a summary of the issues here:
- The 121 centres entered 2088 women into the trial over 39 months. This amounts to five participants from each centre per year. This tiny number opens the term breech trial to potential selection bias.
- Despite being a ‘planned breech trial’, only 21.5 per cent underwent an attempt at external cephalic version (ECV). This would appear to be a rather low proportion, placing a question mark over the adequacy of counselling of the participants.7
- How rigorous and robust was the inclusion process? Despite the TBT inclusion criteria of singleton, non-footling breech with flexed attitude, weighing less than 4kg, the trial reports enrolment of two dead babies, two sets of twins, and anencephalic and a spina bifida baby. In addition, 5.8 per cent of fetuses in the vaginal delivery group were over 4kg at birth, and in 4.1 per cent the type of breech was not recorded.
- Of the women enrolled in the trial, 40 per cent entered with no ultrasound assessment of fetal weight or attitude of the head.
- Of the 13 deaths attributed to the vaginal delivery group: two were ‘most likely’ dead before randomisation; two died after discharge (one attributed to sudden infant death syndrome and the other to gastroenteritis); two died because of respiratory difficulties after birth (calling into question the adequacy of the neonatal resuscitation); one most likely had a congenital anomaly; and a further three had an abnormal fetal heart tracing, but do not appear to have been delivered by caesarean section in a timely manner. Only three died after what was described as a ‘difficult delivery’. In his own analysis of the original data, Glezerman assumed that up to five deaths in the vaginal delivery arm could be attributed in some part to mode of delivery, as against two in the caesarean arm.6 This made the delivery mode PNMR 5/1038 vs 2/1038, yielding a non-adjusted p value of 0.45: a non-significant difference.
- Despite the undertaking that only a qualified person would attend the delivery, 6.7 per cent of the vaginal breech deliveries were delivered by people with little or no expertise as opposed to 2.7 per cent in the caesarean arm. Over 30 per cent of the morbidity/mortality in the group delivered vaginally was from this 6.7 per cent of deliveries. Kierse provides a complete reanalysis of morbidity data.7
- The morbidity data is inconsistent. Fourteen babies in the caesarean arm were said to suffer severe morbidity, but 16 were admitted to a neonatal intensive care unit (NICU). Of more concern, eight of the 39 babies who had ‘severe morbidity’ in the vaginal delivery arm apparently were never admitted to a NICU.
- A skull fracture was sustained during a caesarean birth and a death in the caesarean section arm occurred after the fetus was allowed to labour and deliver vaginally following augmentation with oxytocin. A further 27 per cent of the caesarean section group delivered vaginally after labour was augmented with oxytocin. Incidents and violations of protocol such as these are to be expected in any large, multicentre trial, but bring into question the abilities and commitment of trial participants.
- Of the 69 cases of mortality and morbidity in the original trial, Glezerman could find only 16 that could be attributed to mode of delivery from the data as published. The vaginal delivery versus caesarean section perinatal morbidity rates were calculated as 11/1038 vs 5/1038, with a p value of 0.2, which is not statistically significant.6
- Outcome data presented from the neonates was only short term. Long-term morbidity was not assessed in the original TBT paper.
Despite all these issues, the authors of the trial indicated that perinatal mortality, neonatal mortality or serious neonatal morbidity (as a composite outcome) was significantly lower in the planned caesarean section group compared to the planned vaginal delivery group (RR 0.33; 95 per cent CI 0.19-0.56: p < 0.0001).1
In a follow-up paper published in 2004 (a paper that obstetricians the world over have chosen to either dismiss or ignore3), data from the two-year follow up of infants delivered in the original term breech trial were presented.2 A total of 923 of 1159 children (79.6 per cent) from 85 centres were followed to two years of age. The risk of death or neuro-developmental delay was no different for the planned cesarean than for the planned vaginal birth groups (14 children [3.1 per cent] vs 13 children [2.8 per cent]; RR 1.09; 95 per cent CI, 0.52- 2.30; P = 0.85; risk difference, +0.3 per cent; 95 per cent CI, -1.9 per cent, +2.4 per cent).2 These outcomes suggest that the markers of severe morbidity used in the trial were ultimately not useful.
Further data published have looked at various risk factors in the vaginal delivery arm. Neonatal morbidity is more likely with the use of oxytocics, prolonged second stage and where the birth weight was less than 2.8kg.8 That morbidity was less likely with an experienced clinician.8 Such findings support the Royal College of Obstetricians and Gynaecologists (RCOG) breech delivery guidance.9 The lowest absolute risk for the mother is by way of vaginal delivery. For those delivered by caesarean section the risk is highest if it is performed after labour has commenced (OR 3.33; 95 per cent CI:1.75-6.33, p <0.001).10 It is worth remembering that caesarean delivery for breech may also result in maternal mortality.5
Many of us believe that we were subjected to vehement and at times intensely personal abuse because we dared to question the conduct and findings of the TBT. The final outcomes that were more reassuring are of little comfort. Breech delivery does carry a risk and should be carried out only after careful consideration and counselling.9,11 The art of delivery has in some places been lost through the headlong dash of many obstetricians to the comfort of caesarean delivery and the exhortations of the vociferous randomised controlled trial lobby.
We have been told over the last ten years that the randomised control trial is the only form of trial to be undertaken, and for any other form of trial, ‘we’d suggest you stop reading’.5 Others would disagree with this. A trial is only as good as the relevance of the clinical question it seeks to answer and the parameters it uses to measure outcome. Complex clinical issues like breech do not lend themselves to controlled trials and inappropriate outcomes such as neonatal hypotonia lead to unjustified conclusions.4
To summarise, the term breech trial did not show any benefit for the fetus delivered by a planned caesarean section. Rather, it showed that caesarean section increased maternal morbidity.2,10 Planned vaginal delivery, therefore, must become part of the armamentarium of the competent obstetrician. With regard to the trial itself, we should learn from the issues it raised. As Grant opined, ‘the term breech trial is an example of a randomised trial that was impeccable as regards its methodological design, but questionable as regards its clinical design.’12