AIMS review of the Big Baby trial

Introduction

There has long been uncertainty over whether there is any benefit to early induction of labour when a baby is suspected to be ‘big’, known as ‘Large for Gestational Age’ (LGA). The Big Baby trial was intended to investigate whether there is benefit to inducing labour at 38 weeks in this situation. It was not looking at induction for suspected LGA babies when the mother has diabetes, in which case the risks are different.

Here AIMS critiques the trial and its authors’ conclusions.

Background

The definition of what counts as LGA varies. It can be that a baby has a predicted birth weight over 4kg or over 4.5kg, and/or that the predicted birth weight is ‘over the 90th centile’, which means in the heaviest 10% (one in ten) of the population.

There is a concern that a higher birth weight baby may be more likely to experience shoulder dystocia, which is the situation where the baby’s head has been born but the shoulders do not follow quickly. There is no clear definition of what counts as shoulder dystocia and the figures are not collected routinely so it’s hard to know how common it is. Large scale studies have reported that it affects between about six and seven in 1000 vaginal births¹.

Though shoulder dystocia may be more common in larger babies, the ‘Green-top Guideline’ published by the Royal College of Obstetricians and Gynaecologists in 2012¹ commented that the size of the baby “is not a good predictor” of the chance of shoulder dystocia. The Guideline says this because most large babies are born without any difficulty and many cases of shoulder dystocia occur in babies that weigh less than 4kg and because it is very hard to tell with any accuracy how big a baby is going to be at birth.

The authors of the NICE guideline ‘Inducing Labour’² concluded in 2021 that there was not enough evidence to recommend offering induction for a suspected LGA baby. Instead it recommends discussing with women the options of expectant management, induction of labour or caesarean birth. Nevertheless, it seems to have become increasingly common for doctors to recommend induction to women carrying a suspected big baby, as the high rate of induction in the ‘standard care’ group in the Big Baby trial³ illustrates (see below).

In most cases of shoulder dystocia the use of simple manoeuvres to free the shoulders result in the baby being born quickly and without suffering harm. In around two in 100 cases of shoulder dystocia the baby has a fracture of the collar-bone or upper arm.³ This can be challenging for parents in the short term, though such injuries usually heal completely. Some babies suffer damage to the nerves in the shoulder (known as brachial plexus injury). Reports of how common this is vary from two to sixteen in 100 cases of shoulder dystocia.¹ Again, most of these babies make a full recovery though some (fewer than one in ten of these cases) will have permanent damage. There is no published data on the risk of more serious outcomes such as brain injury or death in cases of shoulder dystocia, probably because they are fortunately very rare events.

What is not clear is whether inducing labour early if a baby is suspected to be LGA reduces the chance of shoulder dystocia or of injury, and what consequences induction in this situation might have for babies or mothers. There seems to have been no research on whether methods to support physiological birth, such as encouraging mothers to birth in upright positions that increase the diameter of the outlet from the pelvis, would reduce the chance of shoulder dystocia.

The Big Baby trial

It is against this background that the ‘Big Baby trial’ was begun. Its stated aim was to determine “does induction of labour at 38+0 to 38+4 weeks’ gestation, in pregnancies with suspected LGA fetuses, reduce the incidence of shoulder dystocia?”⁴

It’s worth noting that there was no science to using induction at 38+0 to 38+4 as the intervention. It was simply picked to try to ensure a big enough difference in length of pregnancy and birth weight between the induction and ‘standard care’ groups.

This was a randomised controlled trial (RCT) with eligible women assigned to either induction between 38+0 to 38+4 weeks of pregnancy or ‘standard care’. Standard care meant whatever would be the usual practice at a participating hospital for pregnancies with an LGA baby, and as we will see from the results below, this also included a lot of early inductions.

The primary outcome to be measured was the rate of shoulder dystocia. The researchers also planned to analyse three secondary ‘composite measures’:

• the risk of neonatal birth injury including one or both of fractures or brachial plexus injury

• the risk of infant complications related to prematurity including one or more of phototherapy or respiratory support

• the risk of birth injury to the mother including one or more of third and fourth degree perineal tears, vaginal/cervical lacerations, or primary postpartum haemorrhage (heavy blood loss)

What was wrong with this trial?

It is AIMS view that the way in which the results of this trial have been analysed and interpreted is incorrect and misleading. We are shocked that the peer reviewers and editors did not identify the many flaws in it prior to publication.

The trial was stopped early

The data monitoring group advised that the trial be stopped early, with only 2893 - rather than the planned 4000 - women randomised. This was because “with the current incidence of shoulder dystocia in the standard care group, recruitment of 12,884 women would be required to achieve 90% power”. This means that because so few women in the standard care group experienced shoulder dystocia, with a sample of only 4,000 women, the study would not have been able to show whether early induction is effective in reducing the rate of shoulder dystocia. It can be appropriate to stop a trial early if the outcomes for one group are very clearly better than for the other, so that it is thought to be unethical to withhold that treatment from anyone. This was not the case here. The trial was stopped early, not because it was showing a benefit or harm from early induction, but because not enough women in the standard care group were experiencing shoulder dystocia to make the results statistically meaningful without a much larger sample. So, the trial was stopped because it was not going to be able to provide any answers on this issue.

Data Trawling

The stated intention was that “All analyses will be by intention to treat at the time of randomisation.” This means that they would compare outcomes by whether the women were randomly allocated to the group having induction at 38 weeks or the one having ‘standard care’ which is the appropriate way to analyse a randomised controlled trial. This analysis showed no significant difference in rates of shoulder dystocia between the induction and standard care groups.

It seems that when the planned analysis failed to show the hoped for benefit from early induction the research team looked for a way to compare the data that would show some differences, which they called ‘per protocol’ analysis. We call it ‘cherry-picking the data’.

This selective, so-called ‘per protocol’ analysis excluded all women from the standard care group who gave birth (whether through spontaneous or induced labour or by caesarean) before 38+4 weeks. This meant excluding from the analysis the outcomes for 370 women, which was over a quarter of the sample. They have not published any analysis to check whether these women were similar in characteristics to the rest of the group randomised to standard care, and it is likely that they were not. This is not good practice and is in itself enough to invalidate the comparison between the selective analysis groups.

Basis of identifying ‘Big’ babies

The criterion for enrolment was that the baby had a predicted birth weight over the 90th centile of a customised growth chart (GROW chart), based on an ultrasound scan between 35 and 38 weeks of pregnancy. In fact, in both induction and standard care groups, about 60% of the babies had a birth weight under the 90th centile, showing that, as the RCOG Greentop Guideline said, it is very hard to predict how large a baby will be using ultrasound. This finding casts doubt on the whole basis of the trial.

Standard care often included early induction

It’s worth noting that a very high proportion of the standard care group had their labours induced at some point after 38+4 weeks. Presumably this reflects the fact that ‘standard care’ for suspected LGA babies in many of the units was to induce at some point - just not always quite as early as 38 weeks.

It’s stated that for the standard care group “induction of labour before 39+4 weeks' gestation was discouraged unless clinically indicated.” However, 32% of these women had inductions before 39+0 weeks, and another 19% between 39+1 and 39+6 weeks. It seems that the care teams found a lot of “clinical indications” to induce women early.

Findings of the intention to treat analysis

When the data was analysed (as stated in the protocol⁵) by “intention to treat” there was no significant difference in the rate of shoulder dystocia between the groups, births were only six days earlier in the induction group, not the 11 days for which the researchers had aimed, and induction only reduced the average birth weight by 163.6g, not the hoped for 300g.

Induction resulted in no benefit in health outcomes for babies. There was no significant difference in either the risk of neonatal birth injury or the risk of infant complications related to prematurity. In fact, none of the babies in either group suffered a fracture. Four babies in the induction group and two in the standard care group had Brachial Plexus injuries but the difference was not statistically significant. There were no stillbirths, but one neonatal death in each group.

For mothers, the intention to treat analysis showed a slightly lower incidence of the composite outcome (third or fourth degree tears, cervical tears or lacerations or primary postpartum haemorrhage). This was due only to a small difference in the chances of a postpartum haemorrhage (defined as blood loss of over 500 ml at birth). This was 44.8% in the induction group versus 49% in the standard care group. These rates seem shockingly high (perhaps because there were very high induction rates in both groups). However there was no difference in the chances of a severe haemorrhage (defined as blood loss of over 1000 ml at birth).

Findings of their selective analysis

As explained above, there are serious issues with this analysis, and we wonder how many different ways they had analysed the data before finding a way that showed some differences. By doing it the authors were able to find a 38% reduction in the rate of shoulder dystocia with induction, from 3.7% to 2.3% of mothers. Note that they were originally looking for induction to reduce the rate of shoulder dystocia by 50% to be considered “clinically worthwhile”.

If the figure of avoiding shoulder dystocia for 1.4% of women is taken at face value that would mean that it would be necessary to carry out 71 inductions before 38+4 weeks to avoid one shoulder dystocia. There would be no benefit for the other 70 mothers and babies.

The authors appear not to have looked at secondary outcomes for mothers or babies in their selective analysis. This means that even if the apparent reduction of one shoulder dystocia for 71 inductions is real, we do not know whether it would be achieved at the expense of increasing other harms. This is a serious gap in the analysis.

The fact that the authors do not comment on whether this analysis demonstrated any health benefits for mothers or babies leads us to suspect that it did not.

What should we conclude from the trial?

The obvious conclusion, based on the intention to treat analysis, is that this trial has not demonstrated any justification for offering induction between 38+0 to 38+4 weeks for LGA babies in the absence of diabetes. This means we still do not know whether or when it might be helpful to offer induction.

The protocol said that induction would be “considered a worthwhile intervention” if it reduced the rate of shoulder dystocia by 50% from 4% in the standard care group to 2% in the induction group. The most they were able to show with the flawed selective analysis was a reduction of 38% from 3.7% to 2.3%. Even if it was true, would this small reduction make early induction “a worthwhile intervention”?

The finding that around 60% of the babies were not in fact over the 90th centile suggests that using ultrasound to estimate whether a baby is LGA is likely to lead to many women being subjected to unnecessary anxiety and potentially to an unnecessary induction of labour, and to many babies being born unnecessarily early.

In relation to the low rates of nerve damage or fractures seen in babies in either group, the authors speculate that “Improved multidisciplinary training on UK labour wards in the decade or so before commencement of the trial might have reduced risks for such complications.” If this is so, is shoulder dystocia really a major issue any more? It would be helpful if the incidence of such injuries in the presence or absence of shoulder dystocia were to be audited.

The protocol said “if we find a reduction in the incidence of shoulder dystocia, but no differences in infant wellbeing outcomes and harm on one or more maternal outcome(s) women and clinicians might here conclude that induction should not be recommended in spite of a positive effect on the primary outcome.”

In the intention to treat (ITT) analysis they found no difference in shoulder dystocia, and no reduction in the risk of birth injuries or prematurity-associated problems following induction at 38 weeks.

Conclusions

For the reasons listed above we question the validity of the finding of a small reduction in shoulder dystocia in their selective analysis. It should not be portrayed to women as evidence that, as the authors claim “compared with delivery at 39 weeks' gestation or later, earlier delivery can reduce the risk of shoulder dystocia.”

In any case, would a small reduction in the chance of shoulder dystocia, in the absence of any improvement in outcomes for babies, really justify inducing labour in all mothers with a suspected big baby at 38 weeks - especially since most of the babies turned out not to be LGA at all?

As is so often the case with our highly medicalised maternity system, the response to a problem such as shoulder dystocia seems to be to reach for a medical solution - early induction. AIMS would like to see research exploring whether a physiology-informed approach to birthing positions would reduce the risk of the problem occurring.

References

1. RCOG Green-top Guideline No 42 ‘Shoulder Dystocia’ March 2012 www.rcog.org.uk/media/ewgpnmio/gtg_42.pdf

2. NICE guideline ‘Inducing labour’ November 2021 www.nice.org.uk/guidance/ng207

3. Boulvain M. et al Cochrane review: “Induction of labour at or near term for suspected fetal macrosomia” 2016 www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD000938.pub3/full

4. Gardosi J. et al “Induction of labour versus standard care to prevent shoulder dystocia in fetuses suspected to be large for gestational age in the UK (the Big Baby trial): a multicentre, open-label, randomised controlled trial” The Lancet, Volume 405, Issue 10491 p1743-1756 May 17, 2025 www.thelancet.com/journals/lancet/article/PIIS0140-6736(25)00162-X/fulltext

5. Ewingtom L.J. et al “Induction of labour for predicted macrosomia: study protocol for the ‘Big Baby’ randomised controlled trial” BMJ Open 2022;12:e058176. bmjopen.bmj.com/content/bmjopen/12/11/e058176.full.pdf