38 weeks is too early

In New South Wales (the part of Australia around Sydney) their public health databases record not just gestational age, but the mode of delivery and whether there was an induction of labour, or labour prior to caesarean delivery. In the same part of Australia, every 3 years, during their first year of full-time schooling, children are evaluated by their teachers using a validated tool called the AEDC, which stands for Australian Early Development Census. This tool evaluates development in 5 domains: physical health and well-being, language and cognitive skills, social competence, emotional maturity, and communication skills and general knowledge.These data are also entered into a database.

Just imagine what you could do if you put the 2 databases side by side!

Well, you could do something amazing like this: Bentley JP, et al. Planned Birth Before 39 Weeks and Child Development: A Population-Based Study. Pediatrics. 2016. You could look at early term, and late preterm deliveries, and find out if having a labour induction at 37 or 38 weeks was safe, in terms of developmental outcomes.

You would have to decide what you considered to be an adverse outcome: such as being below the 10th percentile for 2 of the AEDC domains, which you would then call “developmental high risk” or DHR. This was indeed the primary outcome variable for this study, that looked at gestational age in completed weeks from 32 weeks onward, and mode of, and indication for, delivery. The authors were able to link the AEDC data from 2009 and 2012 to birth data for over 150,000 children.

Nearly 10% of all the children were considered to be developmental high-risk, DHR, and the more immature the child was at birth the higher the relative risk of being DHR, which was true up to, and including, 38 weeks. At 38 weeks there was a 6% increase in the risk of a child being DHR compared to being delivered at 40 weeks.

In other studies of late preterm births, or early term births, the reason a child was delivered early were usually unknown, which has always complicated the interpretation of the results, it was never clear if the reason which led to the early delivery was the problem, or just being born early.

This study partially addresses this, but it still remains a little uncertain why labour was induced at 38 weeks rather than later. Some of the differences that they found may be due to, for example, mothers with early or established pre-eclampsia being induced at 38 weeks, or mothers with a baby showing early signs of growth restriction. the authors have tried to address this by adjusting the relative risks for maternal hypertension and for infants being small for gestational age, which is about as good as you could do with this kind of data.


You can see from this figure, which shows the adjusted relative risks of having DHR, that there really isn’t any difference between 39 weeks, and 40+ weeks. Once you are 38 weeks or less, there are more and more children with DHR, and at each week of gestation, spontaneous labour is associated with lower risks than induction, and the highest risk is actually having an induced labour and then ending up with a C/section.

I think it is highly unlikely that the teachers were aware of the birth history of the children, which makes this assessment practically masked, and I think a very reliable evaluation of the associations between delivery and these outcomes. Of course you can never, from observational data, ascribe causation, but it is hard to think of any other reason why you would find these associations after adjustment, other than a causative link; meaning that it seems most likely that inducing labour, or doing a caesarean delivery before 39 completed weeks, interrupts cerebral development and has long-term adverse effects.

It is also interesting that, even at 40 weeks, there seem to be some risks from having an induction or from a pre-labour caesarean. Although on an individual basis the risks are small, on a population basis this is important, and this information should certainly be included in any shared decision-making about timing of induction or elective caesarean delivery.

The lesson I think from all of this, is that you should have a really good reason for inducing labour, or performing a pre-labour caesarean. The best outcomes seem to be, even if a caesarean is planned, among babies delivered after the onset of spontaneous labour.


Posted in Neonatal Research | Tagged | Leave a comment

Re-evaluating what is really friendly to babies

Many hospitals are very proud of being “baby friendly” or are trying to achieve “baby friendly hospital” certification, under the WHO initiative.

Unfortunately many parts of that program are not evidence-based, such as (as one example) the ban on soothers (or dummies or pacifiers or whatever they are called in your part of the world). But does the baby-friendly initiative even work? Does introducing “baby-friendly” hospital initiatives (BHFI) and/or becoming certified actually improve breastfeeding initiation rates, or duration, or exclusivity?

A new systematic review, punctiliously performed as always by the USPSTF (US public service task force), actually runs to 150 pages in its original version. Fortunately a much more manageable summary of the evidence has been published in JAMA. The review was not just about the BHFI, but a review of all the interventions that have been studied to try and improve the 3 elements of breastfeeding success that I mentioned (initiation, duration, exclusivity).

Overall, the evidence shows that system-wide interventions (such as getting your hospital certified as baby-friendly) have not been proven to improve anything about breastfeeding. The evidence is rather scanty, but what exists is negative (by which I mean not that the evidence is of a negative impact, but that there is no evidence of benefit). Various individual parts of the BHFI, as system-wide approaches, have also not been shown to be  effective, either.

The system-level interventions that were evaluated within these 9 studies included receiving accreditation for the BFHI, a clinic policy to provide breastfeeding support groups for pregnant women and breastfeeding mothers, and establishing maternity care practices for maintaining mother and infant contact following delivery or restricting or delaying pacifier use. Across these 9 studies (7 RCTs and 2 before-after studies), there was no consistent evidence of an association between system-level changes and the rate of any or exclusive breastfeeding at up to 16 weeks’ postpartum.

In contrast, individual-level interventions, which include a whole host of different interventions such as individualized lactation support, educational interventions, peer counseling, and telephone support, were effective, and were maybe more effective if applied at multiple time points, antenatally and postnatally.

One of the individual level interventions which did not work in 2 RCTs was counseling to avoid pacifier use.

An editorial accompanying the USPSTF papers gives a clear interpretation of the review and the recommendations that the USPSTF makes based on them. It also includes this:

A second potentially controversial area involves use of feedings other than breast milk. Counseling mothers to avoid giving infants any food or drink other than breast milk during the newborn period is step 6 of the BFHI and one of the primary care interventions most commonly used to support breastfeeding. Three randomized trials have specifically examined the effectiveness of counseling to avoid giving newborns any food or drink other than breast milk; none showed a beneficial effect of such counseling on breastfeeding duration.

The editorial notes that these 3 studies were not included in the new SR. I checked over the inclusion criteria for the USPSTF review, and I don’t really understand why they weren’t included. It was maybe because it is an intervention that only applies to mothers who are, or were intending to, breastfeed, but that isn’t listed as an exclusion criterion in the methodology part of the 150 page document.

Avoiding any and all supplemental food or drink is potentially hazardous. A mother’s milk may take several days to “come in” and babies can get dehydrated with hyperbilirubinemia, hypernatremia (occasionally severe), and may need to be hospitalized. Those complications are strongly associated with exclusive breastfeeding. Such complications might be acceptable if there were proven adverse consequences of occasional supplementation of breastfeeding babies with formula if there is an indication to do so. But I don’t think that is proven, at all. Nor is there an evidence in this whole systematic review of impacts of the various interventions on health outcomes. Even though the individual level studies do mostly show benefits on initiation of breastfeeding, breastfeeding for at least 3 months, and exclusive breastfeeding for at least 3 months, there is no clear decrease in gastroenteritis, or respiratory infections. Which is not too surprising as few studies have actually measured those outcomes, and the observational studies showing those impacts of breastfeeding often show relatively small effect sizes in the high-resource countries where many of these trials have been done. You would really need a very large trial in a higher risk population to prove, what I think there is little doubt about, that a specific intervention which actually increases breastfeeding rates and durations will have significant health benefits.

It is because of those benefits that we need better data. Not data about the health benefits of breastfeeding, they are already overwhelming. Data about how to improve breastfeeding initiation and duration, better studies to prove what works so that we can focus our resources on effective and beneficial practices.

The avoidance of soothers/pacifiers also came up in another influential review, this time from the AAP, and about reducing sudden unexpected death in infancy, which recommends considering pacifier use at sleep times. In the text they specify that this should only be after establishment of breastfeeding, but what they mean by “well-established” is not clear, nor do they give any data that shows that early pacifier use adversely impacts on establishment of breastfeeding.

Improving breastfeeding duration requires us to be supportive of mothers who are trying their best without the extensive social supports that societies provided in the past, without a network of wet-nurses, mothers who may be surrounded by ill-informed healthcare workers, family, or friends. A mother who feels good about herself and her attempts to give breast milk to her baby, as much as she can for 3 or 4 months if she can, will be much more likely to be successful than if she is made to feel inadequate by policies or by individuals that demand perfectly exclusive breastfeeding for 6 months.

Let her give the baby a soother/pacifier/dummy if she wants, and if the feeding is not going well in the first few days, a few bottles of formula, or pumped breast milk are not only harmless, they are better than harmless, they may help the mother and her baby to get over the hump and carry on breastfeeding, rather than giving up all together. They may even get the Dad more involved, many of us like being involved in feeding babies, and feel a bit jealous when breastfeeding is going well and we are excluded!

That sounds really baby friendly to me.

Posted in Clinical Practice Guidelines, Neonatal Research | Tagged , | 3 Comments

Antenatal Steroids closer to term?

A new systematic review in the BMJ (Saccone G, Berghella V. Antenatal corticosteroids for maturity of term or near term fetuses: systematic review and meta-analysis of randomized controlled trials. BMJ. 2016;355:i5044) includes 3 studies which examined the use of antenatal steroids in mothers at high risk of delivery between 34 and 36 weeks 6 days, and 3 studies of using antenatal steroids prior to elective section after 36 weeks.image1

The results are firstly presented with both categories combined, which I think is very questionable. I can see no clinical value in combining the 2 categories of studies, there is no overlap in the gestational ages, so the combined outcomes are meaningless. No mother could have been eligible for both of the groups of trials, which I just realized is a good criterion for deciding if you should perform meta-analysis of trials.

Barrington’s new rule for meta-analysis: If it is impossible to have been eligible for both of 2 trials, don’t combine them. Now I will have to go through my own Cochrane reviews and see if I follow my rule.

The reasoning behind the rule should be obvious, the only point of doing a Systematic Review (SR) of an intervention is to be able to answer a clinical question. Such as “if I have a baby at term with hypoxic respiratory failure should I use inhaled nitric oxide, what are the benefits, and what are the risks, compared to no nitric oxide?” So meta-analysing all the studies which randomized babies for whom that definition is accurate is appropriate.

The first 3 forest plots in the antenatal steroid SR on the other hand give an answer to a question which will be never be asked, “if I have a mother who is somewhere between 34 and 40 weeks gestation and might be high risk of delivering before 37 weeks or on the other hand might be planned for an elective caesarean delivery after 37 weeks, should I give her steroids at some point, either now or….” (I gave up trying to formulate the question).

The SR does, thankfully, go on to also provide the results of the 2 groups of trials separately, here is an edited version of table 3, which is the results of neonatal outcomes for the studies of mothers at high risk of delivery between 34 weeks and 36 6/7.

The first fractions are the results from the steroid group, the second are the controls. As you can see it really is only the multi-center MFM network trial which contributes much to the results. Don’t ask me what the column headed τ2 is, it isn’t explained in the methods.

The results whose 95% CIs do not include 1.0 (no impact of antenatal steroids) are in bold. The summary measures are all relative risks, apart from the mean reduction in Apgar scores by 0.62 (I don’t recall ever seeing a baby with an Apgar score of 7.98!)

Although 40 % less severe RDS and 40% less need for surfactant sounds great, another failing of this SR is the complete lack of any calculation of absolute risk reduction or NNTs, so I have done it for you. The absolute risk reduction is from 2.4% to 1% in the Gyamfi-Bannerman trial, an ARR of .014, or a number needed to treat to prevent 1 case of severe RDS of 71 (I can’t calculate the 95% CI without putting all the numbers into another computer program), which is a lot of injections for no decrease in need for intubation, but avoiding surfactant in one baby.

The downside, as you can see from that table, is a substantial increase in the incidence of neonatal hypoglycemia, from 15% to 24%. an absolute risk increase of 0.09, with a number needed to harm of 11. Hypoglycemia is defined in the supplementary material of Gyamfi-Bannerman as a glucose at any time of less than 40 mg/dl, which is 2.2 mmol/L for those of us who use modern units. The protocol doesn’t state anything about standardized monitoring for blood sugars in the babies of the participants that I can see (the protocol is available on the NEJM website, and I read it, just to be able to write this blog post for you my gentle readers), as it was a masked study we must assume that the babies were monitored in the same way in the 2 groups, so the increase is probably a real effect of the intervention. As you can see from the first table, mothers with diabetes were excluded from this trial.

One conclusion of this analysis is therefore that there is a substantial increase in the risk of hypoglycemia, and therefore all babies born after antenatal steroids will need to be monitored for this effect, in the same way as babies with other risk factors, which is a lot of extra blood sugar monitoring and extra pain, for 100% of the babies, for a non-significant decrease in the number of babies needing mechanical ventilation of 0.6%!

I am actually quite surprised at the BMJ, who have one of the highest standards of editorial review, in my opinion, that this SR was published like this.

If we go on to the issue of steroids before elective caesarean delivery, it gets even murkier.


There is no “neonatal hypoglycemia” in this table, because it wasn’t reported in any of the trials. I think we have to assume therefore that there might well be a increase in the incidence of hypoglycemia for this group of babies also, and treat antenatal steroids as a risk factor for hypoglycemia in all of the babies.

The supposed benefits of antenatal steroids are, for example a fantastic 78% reduction in severe RDS. Wow! But the reduction in severe RDS is, in absolute terms, from about 1% to about 0.25% so a number needed to treat, NNT, of over 100.

The 2% reduction in NICU admission (from 5% to 3%) gives an NNT for admission to the NICU of about 50. Some of the outcome variables are pretty dumb, 2 hours less on oxygen for example, only applies to the babies who needed oxygen to start with, and is of no real clinical importance.

As you can see from the table, in fact, the rate of admission to the NICU was not different. Which is not at all how the article by Stutchfield was actually originally reported. They claim to have a reduction of 50% of NICU admissions, but in other places in their publication they note that was a reduction in NICU admissions “for respiratory distress”, in fact the total NICU admissions were NOT different between groups, as admissions for non-respiratory disease were in the other direction, more among steroid exposed babies than controls. Maybe they were for hypoglycemia? The numbers they report for NICU admission by gestational age completed weeks, are actually just for respiratory distress, and are taken from logistic regression, not from the original data:

The predicted probability of admission at 37 weeks was 11.4% in the control group and 5.2% in the treatment group, at 38 weeks it was 6.2% and 2.8%, respectively, and at 39 weeks it was 1.5% and 0.6%.

Which means that at 39 weeks the NNT is about 111 to avoid admission to the NICU for respiratory distress, but overall admissions are not actually significantly reduced.

I do think that avoiding NICU admission is a worthwhile goal, and reducing neonatal respiratory distress is also worthwhile. But with such a small absolute reduction in these outcomes we had better be sure that the intervention is safe. Only one of these trials has tried to do long term follow up of the babies. Stutchfield sent out a questionnaire to parents of babies born in 4 of the original centers in the trial (n=about 800), only about half were returned, and about 7/8 of those also had a school questionnaire returned.

The abstract of that manuscript states:

25 (12%) children whose mothers received betamethasone had reported learning difficulties compared with 27 (14%) control children. The proportion of children who achieved standard assessment tests KS2 exams level 4 or above for mathematics, English or science was similar.

The actual data presented in the body of the article don’t report the proportion achieving those levels, which is weird, but the article does report other outcomes which are not as comforting:


With the large number of comparisons made this might be a spurious finding, but it certainly should give some pause before recommending antenatal steroids for this indication, a recommendation which might apply to many thousands of mothers and babies every year.

So what are current recommendations?

Well there are some really really new ones, in October’s Obstetrics and Gynecology the American College of Obstetrics and Gynecology have published an update to their antenatal steroid guideline.  It now states:

A single course of betamethasone is recommended for pregnant women between 34 0/7 weeks and 36 6/7 weeks of gestation at risk of preterm birth within 7 days, and who have not received a previous course of antenatal corticosteroids.

As far as I can see there is nothing about steroids for elective caesarean delivery near to term. Apart from it being in the list of things for which there are no data(!)

The Royal College of Obstetrics and Gynecology of the UK in contrast in their statement from 2010 note the following:

Antenatal corticosteroids should  be  given  to  all  women  for  whom  an  elective  caesarean  section  is planned prior to 38 +6 weeks of gestation.

I think these two guidelines are both mistaken, neither of them has bothered to look at the absolute risk reduction (or increase), the number needed to treat, or the potential harms. I think the recommendations should be much more nuanced, and note the very large number of treated women for whom there will be no benefit, that the decision should/must be discussed with the mother (who knows maybe even the father).

I think the biggest potential adverse effect of these guidelines is that the risks to the baby of a late preterm delivery will be minimized; that obstetricians, who already think that late preterm birth is unimportant in the long term, will think that they can blithely deliver babies before term as long as they give steroids.

We need to emphasize that late preterm delivery is a major risk factor, on a population basis, for adverse developmental and health issues in children.

A great review article from Betty Vohr in 2013 laid out all of the then available data showing adverse impacts of being born before 37 weeks, we now, of course, have even more data, and more reasons for concern about these infants. Another review from last year, with Elaine Boyle as first author also noted the adverse long term physical health effects of being late preterm or early term, compared to really really full term.

Including this figure for example:


The impacts of late preterm, or even early term delivery should not be minimized just because we might be able to slightly reduce the risk of some short term respiratory adverse effects with antenatal steroids. Only when preterm delivery is inevitable or essential for the mother or baby should we start to weight the risk/benefit balance of antenatal steroids, which doesn’t seem to me to be a closed question: with total NICU admissions and need for mechanical ventilation unchanged, and substantially more hypoglycemia.

Posted in Neonatal Research | Tagged , , , | 2 Comments

What outcomes matter to parents? A new publication

This new publication of ours has been an interesting process, Annie Janvier and I wrote it in collaboration with other parent representatives, Barb Farlow, who we have collaborated with previously, a mother of a little girl who had trisomy 13, and Rebecca Pearce and Jason Baardsnes, who had twins at 25 weeks, one of whom died. As many of you will know, Annie and I are also parent “representatives”, our little treasure Violette (whose hand is at the top of this blog) was born at 24 weeks gestation.

Janvier A, Farlow B, Baardsnes J, Pearce R, Barrington KJ. Measuring and communicating meaningful outcomes in neonatology: A family perspective. Semin Perinatol. 2016 In this article we discuss the outcomes that are commonly measured in neonatology, both in research, and in quality control, and for individual assessments, and ask if we are measuring the right things.

We wanted to be thoughtful and provocative (wanting to provoke change, not just strong reactions), and I think we got it right, let me know if you agree:

Dichotomous and continuous outcomes : we discuss how things which are in reality continuous outcomes, such as lung injury, or developmental delay, have been dichotomized into yes/no, good/bad outcomes, mostly for the convenience of researchers. It makes little sense to label someone who stays in oxygen until 36 weeks and 2 days as having BPD, whereas if the oxygen was weaned 3 days earlier they would not have had BPD. What is important is the impact of that lung injury on the life of the baby.

Even more problematic is combining artificially dichotomized outcomes especially combining them with death (death or BPD, for example) as if coming out of oxygen later were in some way equivalent to being dead.

Of course combining multiple artificially dichotomized outcomes is even more problematic, death or NDI being the most important example, being dead, or having a serious hearing deficit requiring hearing aids, or having a major visual problem, or being stiff on one side, or having a developmental delay which puts you below -2 SD of the population mean, are often combined, as if it was self evident that each one of those adverse outcomes were equally important. We note :

When two outcomes have potentially very different values to patients, they should be discussed separately.

So outcome data combining outcomes which are of very different individual value are of little use. Discussing the risk of death separately to the potential long-term consequences if the infant survives, may give a very different response, to the “80% risk of death or NDI”.

In the article we also discuss other important issues such as “Confusing screening tests with disabilities”, the Bayley Scales of Infant Development should be considered to be a screening test, a way of evaluating current development, screening for delays, and plugging kids into intervention programs if necessary, but NOT used as a way of defining who is OK, and who has an impairment; we know the Bayleys are very poor at prediction of longer term intellectual problems, and I know of no data that low Bayley scores are related to poor quality of life among NICU graduates.

Some outcomes which have a huge impact on families are, in contrast, minimized in neonatal outcome studies: feeding difficulties and need for home gavage feeding for example, are rarely quantified in studies but may be very disruptive for families.

We also discuss the importance of resilience, or “coping”, how many families can adapt to non-optimal outcomes, and can even benefit from them, how resilience has been undervalued in neonatal research:

… a family with a child who has severe cerebral palsy, for example, will find meaning and hope in the life of their infant, even if they wish it was otherwise. The same can be said about families of children who have home ventilation, Down syndrome trisomy 13 or 18. [for each of these examples we give references] Adaptation and coping have been measured in older patients, but neonatal research is lagging behind. Several studies have noted that the adverse impact of a child with neurological or developmental difficulties decreases markedly over the years. A recent study of families with children who have intellectual disabilities was almost universally positive about the impacts of having such a child on themselves and their family. Interventions aimed at families with poor resilience predictors could improve neonatal and parental outcomes.

We end the article with some recommendations which I reproduce below in shortened form.

Recommendations for Clinicians

1. The caregiver should themselves be informed regarding the condition, the risks, and the potential impacts of the condition for children.

2. Avoid conflating neonatal outcomes: Remember that for a parent, death is not the same as disability. Clinicians should speak about the risk of death, and then of disability in survivors.

3. Children are more than a diagnosis : Parents should not only be informed of potential adverse conditions or diagnoses. They should know what these diagnoses mean to children and families, in a practical sense.

4. Personalize the information: The information should be adapted to the needs and wishes of the family, and the precise nature of the infant’s situation. Many recommendations for antenatal counseling in extreme prematurity demand that information be standardized and that providers describe all the possible complications of prematurity.

5. Provide balanced information: Parents should be given an opportunity to ponder what a diagnosis means for them, for their family, for their future, they should be told about the positives as well as the negatives.

6. Provide information relative to follow-up programs : Parents are often not well informed of the reasons for systematic follow-up. Many parents think this is essential to identify potential problems their child may have…

7. Discuss real-life outcomes: Providers should have regular interactions with NICU parents about the life trajectory of their child: what is the next step—remove the tube, remove the CPAP, full feeds, transition to oral feeds, oximetry, etc…

8. Normality and normativity: Avoid the term “normal” when describing babies after NICU discharge. Their tests and imaging may be “normal” or “abnormal,” above or below average; but for a family—especially after the NICU experience—these babies are extraordinary.

9. Empower: Parents of all NICU children should be able to feel like good parents—doing a good job caring for their child. In the NICU, they should be congratulated when they hold their baby, visit, provide breast milk, read to their child, ask questions, etc. .

10. Inform NICU families of parental support groups or associations that can offer them experiential information about the child/family experience

11. Words are important : Know the name of the baby and avoid labels: neonates are not their condition or their outcome. The way we describe babies can have an impact on how parents see their children.

Recommendations for researchers

1. Research projects should focus on primary outcomes that are of importance to families. Families should be involved in planning research, and in determining which outcomes should be investigated…

2.Decreasing the impacts of neonatal critical illness is a vitally important task, given the improved survival of infants with so many different conditions. But unless we investigate which outcomes are truly most important to our patients and their families, we will not be able to progress further.

Posted in Neonatal Research | Tagged , | 2 Comments

Delivery room interventions for the profoundly immature: what are we doing, what should we do?

I still haven’t found the best term for babies who are born so immature that their chances of survival are significantly reduced. Extremely Low Gestational Age Newborn, or ELGAN is a term which is not pretty but has the advantage of being accurate and not being associated with any value judgement, it is, however, now very strongly associated with Dr Alan Leviton (and many others)’s cohort study, almost becoming a trade mark! “Threshold of viability”, (or infants born at the…) has the disadvantage of being a moving target and not being clear as to how viability is determined; as we will soon see, if you decide a baby in this gestational age range is not viable, they will die. “The periviable period” is, I think a reasonable term for the time between 20 and 24 weeks, but I really don’t like “periviable neonates” for the same reasons I just mentioned, and “infants born during the periviable period” is unwieldy. Maybe “profoundly immature” is a better term, no values involved, and clear to everyone what we are talking about…

Two very recent publications and their accompanying editorials are worth discussing. They have different methods and conclusions as you will see:

The first is a study from France, part of the EPIPAGE-2 study. Perlbarg J, et al. Delivery room management of extremely preterm infants: the EPIPAGE-2 study. Archives of disease in childhood Fetal and neonatal edition. 2016.

This study was a high quality prospective cohort study of very preterm deliveries in all but one of the administrative regions of France. To clarify, France is divided into 22 metropolitan regions (that is, mainland France) and 4 overseas regions (Guadeloupe, Martinique, La Réunion, et Guyane (French Guyana)), or at least it was, as the number of regions has been decreased recently, it seems. All but one of the metropolitan regions was included in EPIPAGE-2, and all of the overseas regions. The timing of inclusion overlapped but wasn’t exactly the same for each region, which is a shame; but given the enormous nature of this study we can forgive a few logistic problems.

The protocol publication (Ancel PY, Goffinet F. EPIPAGE 2: a preterm birth cohort in France in 2011. BMC pediatrics. 2014;14:97) notes the following :

EPIPAGE 2 is a population-based prospective study scheduled to follow children up to the age of 12 years. Eligible participants include all infants live born or stillborn and all terminations of pregnancy between 22 and 31 completed weeks of gestation in all the maternity units in 25 French regions (21 of the 22 metropolitan regions and 4 overseas regions) during the inclusion period. The only region that did not participate accounted for 18 415 births in 2011, i.e., 2.2% of all births in France

The babies in the study were born between May and December 2011, in some regions the time periods of data collection were not exactly the same, but they all overlapped.

This particular publication is about what happened in the delivery room for babies who were thought to be between 22 and 26 weeks gestation. It includes just over 2000 births, and is a good description of French practices during that period.

One of the problems with large regional cohort studies like this is that they include very acute unexpected extremely preterm deliveries in small general medical centers, as well as controlled deliveries in tertiary teaching hospital centers. There are many layers of subtlety that are difficult to grasp from afar.

Nevertheless, there are a number of findings worthy of discussion. At 22 weeks gestation there were 421 deliveries, of which 52 were “terminations of pregnancy” we don’t know from the paper what the indications for the termination were; (some “terminations” are performed because of preterm labour, which is quite different to terminations performed because of life-threatening malformations, for example). Of the remaining pregnancies, most led to stillbirth (74%, of the original group, but 85% after excluding terminations of pregnancy) and, among the others, only one of the live born babies received active intervention with oxygen and intubation, that baby died, leading to a survival of zero.

Even stillbirth, however, is not necessarily a hard endpoint; what I mean is that an attitude of active management of profoundly immature delivery will dramatically increase the number of babies born alive, compared to those born without a heart rate. It is also the case that some babies who are classified as stillbirths may be apneic, and, as a decision for comfort care has already been made, no-one seeks a heart rate. They may thus be resuscitatable, but are classified as stillborn.

If we try and contrast these results to the Swedish national cohort (EXPRESS), that study excluded pregnancy terminations, and at 22 weeks gestation there were 142 deliveries with 91 still births or 64% were stillborn, of the 51 liveborn infants there were 23 admitted to the NICU with 6 surviving to discharge.

There are differences between the 2 cohorts at 23 weeks also, in the EXPRESS study there were 183 deliveries and 82 were stillbirths, or 45%; they also report the intrapartum deaths, that is those stillbirths for whom the baby was known to have had a heart rate prior to labour, which were 19, or 10%. There were 16 deaths in the delivery room, leaving 81 admitted to the NICU, with 53 survivors (or 65% of NICU admissions). The EXPRESS study does not give the proportion of babies who had a decision for comfort care before delivery, but almost all of those born at a level 3 hospital had a neonatologist present at birth, and most were intubated.

In the French cohort there were 361 deliveries at 23 weeks that were not terminations, 270 were stillbirths, or 75%. That leaves 89 who were born alive, most did not have active intervention in the delivery room, seven of the babies were admitted to the NICU and 1 survived.

At 24 weeks there are still major differences between the 2 cohorts, many more stillbirths, and delivery room deaths, and NICU deaths, in the French cohort than the Swedish.

At 25 weeks there are smaller differences in the delivery room, but NICU deaths remained higher, 65% survival of NICU admissions at 25 weeks in France compared to 82% in Sweden.

In the French study, they examined what factors were associated with the decision-making regarding withholding or withdrawing intensive care, in addition to gestational age, babies under 600g were more likely than larger babies to have care withheld or withdrawn, babies from IVF were less likely to have this, but there was no difference between boys and girls. We know that smaller babies at any gestational age have lower survival, so to take that into account in decision-making is reasonable, but IVF has no known effect on survival, and sex certainly does, effect at least as important as birth weight. Which points out that decision-making is driven by values and not just by data, even if national guidelines have, until recently, uniquely spoken about the risks of death or disability as the driving force for decision-making.

The authors of the article point out that survival rates to discharge are lower in France, than in Sweden, the USA, Japan or Australia, at 23 up to 26 weeks gestation. Part of which difference is associated with the relative rarity of active intensive care below 25 weeks, but intensive care was given to the majority of babies at 25 and 26 weeks, yet survival rates were still lower. The authors speculate that later withdrawal of intensive care may be more frequent, given how frequent it is before NICU admission in the profoundly immature, and also that places who give intensive care to the most profoundly immature babies have lower mortality for more mature infants. That second explanation is supported by results from the NICHD NRN (Smith PB, et al. Approach to infants born at 22 to 24 weeks’ gestation: relationship to outcomes of more-mature infants. Pediatrics. 2012;129(6):e1508-16), and I think it likely that both are important.

The accompanying editorial Janvier A, Lantos J. Delivery room practices for extremely preterm infants: the harms of the gestational age label. Archives of disease in childhood Fetal and neonatal edition. 2016;101(5):F375-6. points out the same issues and wonders if there is any other place in modern medicine where interventions are universally denied in one high-income country, and almost universally applied in another.

I have asked the same thing about these differences, which also occur even within the USA. Is there any other condition in modern medicine, where one university center has 0% rates of intervention, and another, even in the same network, has 100%? And yet all of us claim to be practicing shared decision-making! I think the share is clearly not equally divided.

The second article is a regional study using administrative databases from New South Wales, Australia. (Haines M, et al. Population-based study shows that resuscitating apparently stillborn extremely preterm babies is associated with poor outcomes. Acta Paediatrica. 2016. Which is the kind of title that makes me sarcastically say “what a surprise!”) Between 1998 and 2011 they examined the outcomes of over 2000 babies of 22 to 27 weeks and 6 days who were born with a 1 minute Apgar of 0.  As far as I can tell, and this still isn’t quite clear to me after re-reading, the babies were classified as stillborn or live-born by the local personnel at the time who filled in the official  documents. The abstract states “We classified 2173 infants….. as stillborn” but I don’t see any methods for the authors to have done that. Of the infants classified as stillborn there were 40 who had an attempt at resuscitation, none were admitted to NICU, one surviving 51 days, but none surviving to discharge. I thought at first when I saw that all the stillborn resuscitated babies had died, that a baby who never responded during resuscitation and never had a heart rate might be considered to be stillborn, I think it’s weird that a baby who lives for 51 days would be called stillborn! I guess when you look at over 2000 records you are bound to find weird things, some of which might just be errors.

We also don’t know from these data which babies had a fetal heart rate detected in the hospital before delivery, and how many of the non-resuscitation babies were already known to have deceased in utero.

Of the 89 infants who were classified as live-born 48 had a resuscitation attempt, 13 were stabilized enough for NICU admission and 11 were discharged alive. The survivors ranged from 23 to 27 weeks gestation.

None of the infants in this study who still had an APGAR recorded as 0 at 5 minutes of age survived.

The conclusion of the study states that even with resuscitation attempted, “almost all of the babies died”. I don’t agree that 11 survivors out of 88 resuscitation attempts is “almost all”. The comparison group, babies with 1 minute Apgar 0 and no resuscitation attempt, had zero survivors.

The data on the lack of response when the babies were still pulseless at 5 minutes is useful I think, it is consistent with a somewhat different data set that we published a few years ago (Janvier A, Barrington KJ. The ethics of neonatal resuscitation at the margins of viability: informed consent and outcomes. The Journal of pediatrics. 2005;147(5):579-85) where we showed that if the babies at 23 weeks needed extensive resuscitation, and still were bradycardic by 3 minutes, they usually died, or may have survived but with major short-term complications. Extending this out to being pulseless at 5 minutes might give us enough certainty for these extremely preterm babies that they aren’t going to make it, and potentially stop resuscitative efforts.

La Gamma EF, et al. Resuscitation of potentially stillborn periviable neonates: who lives, who dies and who gets missed? Acta Paediatr. 2016;105(11):1252-4. The editorial accompanying the Australian study puts the results of the study differently; they state the following : “active resuscitation of ELGANs with a zero Apgar… increases survival”. They suggest that active intervention and a trial of therapy is the preferred approach for babies such as these, but that such an approach should be accompanied by an understanding by the neonatal team (which I hope includes the parents) and when futility is achieved. I think “futility” is not the best term to use here, it is indefinable and slippery. I would prefer that we just talk about frequent re-evaluations of the status of the baby, with an honest re-consideration of the goals of care. Trying to pre-establish limits of interventions or of acceptable complications is helpful but difficult, and also tends to slip away.

My take on these 2 articles is that, as I have said many times before, making a decision and having strict intervention guidelines based on gestational age alone makes no sense. We usually do not know gestational age with accuracy, and even if we did, there is a huge range of potential outcomes at any gestational age, which overlap by much more than 2 weeks. Babies with a significant chance of survival, taking into account all of the prognostic information available, should be offered a trial of therapy, which should include extensive delivery room resuscitation if required. What “significant chance” means is different for different families, and a major part of the antenatal consultation should be involved with establishing a relationship of trust with the parents and exploring their values in order to define what that means for them.

I’d encourage everyone to read also a new “viewpoint article” from 2 members of our team,  Gaucher N, Payot A. Focusing on relationships, not information, respects autonomy during antenatal consultations. Acta Paediatr. 2016.  which discuss the relational aspects of the antenatal consultation, based on research that has been done in Québec, and elsewhere, over the last several years, including the latest publication that I discussed recently on this blog.

Posted in Neonatal Research | 2 Comments

Oxygen, getting the dose right. Not so easy.

Preterm babies require differing concentrations of oxygen to maintain them within the optimal saturation range, which is clearly the low 90’s, from all the data we have so far (Saugstad OD, Aune D. Optimal Oxygenation of Extremely Low Birth Weight Infants: A Meta-Analysis and Systematic Review of the Oxygen Saturation Target Studies. Neonatology. 2013;105(1):55-63.) Of course there are still many unknowns, such as the importance of intermittent hypoxia compared to persistently lower saturations, the importance of intermittent hyperoxia, which is not infrequent and may frequently follow apnea. How important is keeping the saturations stable? I think it probably is quite important, but extremely difficult to know, and difficult to achieve.

How do we make sure that preterm infants always get the dose of oxygen that will keep them in that range?

Many years ago there were publications about the development of servo-controlled oxygen blenders which were linked to transcutaneous oxygen electrodes. They never reached commercial exploitation.

Which has led to the current situation where oxygen saturation alarms are the most frequent alarms in the NICU, alarms which often require manual adjustment of oxygen dose, sometimes hundreds of times a day. There are now several servo-controlled oxygen dosing devices in various stages of development and exploitation.

I think these devices could be an enormous advance in neonatology, but there are a few things that will have to be addressed :

  1. How to ensure that the oxygen isn’t increased when a baby is apneic. If the baby isn’t breathing then increasing oxygen concentrations will be ineffective to improve saturations, and may well lead to fairly prolonged hyperoxia when they start breathing againvan (Zanten HA, et al. The risk for hyperoxaemia after apnoea, bradycardia and hypoxaemia in preterm infants. Archives of Disease in Childhood – Fetal and Neonatal Edition. 2014). On the other hand, there may well be lung volume loss during an apnea and an increase in VQ mismatch, and a temporary increase in O2 needs during recovery (I am not sure this has ever been demonstrated, but it is a possibility) which means that the O2 should probably remain constant during the apnea, but then be programmed to be ready to increase when the baby starts to breathe again.
  2. The linked concern is how we will be sure that post-apneic hyperoxia isn’t worsened by servo-controllers. I think it is likely to be improved, with more rapid reduction to baseline requirements.
  3. How to ensure that periodic breathing isn’t intensified and prolonged by the servo-controller. Periodic breathing is a pattern of breathing that is driven by the peripheral chemoreceptor.  It is maintained by a phase-shift of stimulus (hypoxia) and response (increased respiratory drive consequent on increased chemoreceptor afferent activity), so when you examine recordings of babies with prolonged periodic breathing you will often see (depending on response times of all the elements in the system) that saturations are increasing when the baby is in the apneic phase and decreasing when the baby is breathing (Barrington KJ, Finer NN. Periodic breathing and apnea in preterm infants. Pediatr Res. 1990;27(2):118-21.) This happens because of the delays in the physiologic response system, in other words when a baby stops breathing it takes a while for the blood passing through the lungs to desaturate, even more time for that blood to reach the peripheral chemoreceptor (which is by the way the only part of the respiratory control system that responds directly to oxygen tension) and then more time again for the chemoreceptor responses to be translated into phrenic nerve activity. (and that explanation jumps over several intermediate steps). Periodic breathing can last for hours in some babies, and may be associated with large fluctuations in saturation.  I think there is a real chance that servo-regulated FiO2 could re-inforce these cycles, and might lead to prolonged repetitive desaturation/resaturation events. Which might (or might not) be harmful.
  4. The pulse oximeters will continue to have alarms set, can we use these new systems to make the alarms smarter? In the NICU when a baby in in room air there is no value to having a high saturation alarm, so the high alarm is usually switched off. If the baby has a transient desaturation and the oxygen is increased a little, what often happens is that the baby will recover back to high saturations, and, as the high alarm was switched off, the baby may over-saturate for prolonged periods of time. It shouldn’t be too difficult, once there is a link between the oximeter and the oxygen blender, to switch of the high alarm when the baby is in room air, and switch it back on again when the baby is receiving oxygen.

As a result of these concerns, I think that we do need to prove that servo-regulated oxygen devices improve clinical outcomes, or at least does not worsen them.

The stimulus from this blog post came from a new study published by the neonatal group from Tasmania, (Plottier GK, et al. Clinical evaluation of a novel adaptive algorithm for automated control of oxygen therapy in preterm infants on non-invasive respiratory support. Archives of Disease in Childhood – Fetal and Neonatal Edition. 2016).
using a novel system that they have developed, they studied 10 preterm infants on non-invasive support with a cross-over design, and showed a major reduction in time outside the oximetry target range. Their new system incorporates several potential improvements in the response algorithms.

An accompanying editorial from Christian Poets and Axel Franz is well worth reading also, it includes this interesting graphic which shows, firstly that there are more stories than I was aware of, for each study they show with the small horizontal lines, what the target saturations were, i.e. they were between the light and dark grey lines. Then they show what proportion of time the manual adjustment of FiO2 and the automated control of FiO2 were within the target range.


Some of their interesting thoughts about this situation, you can see that the proportion of time within target range for manual control was extremely variable, these are all small corss-over studies, so it may be differences in the patients that is the reason behind this. All the studies showed more time in target range, and the degree of improvement was quite variable also, with the new study having one of the greatest improvements in percentage in the target range.

The new study also showed very little time with extreme hypoxia (<80%) and very little time with extreme hyperoxia while receiving oxygen, all of which were statistically significantly better than manual adjustment.  The manual care periods were associated with more than 2 adjustments of oxygen of more than 1% per hour on average, the aut0matic by over 60 adjustments of more than 1% and about 600 adjustments per hour of at least 0.5%, which is the way the automated blender is controlled.

As I started off saying, I think this is going to be a major improvement in neonatal care, but as for many other things, we need to have some evidence of improved clinical outcomes, otherwise it will be difficult to get funding for the new equipment, installation and training that will be required.

Fortunately, as Poets and Franz note, there is a trial in the works for which funding has been approved. In the meantime continuing to improve the algorithms will be necessary.


Posted in Neonatal Research | 1 Comment

It’s only mild Encephalopathy; now can we stop worrying?

I learnt as a fellow that infants whose maximum grade of HIE was Sarnat 1, or mild, had normal outcomes. I recognized that the data to support that were limited, but one of the best older studies was by my mentor and friend Neil Finer, who together with Charlene Robertson organised the follow-up at 3.5 years of 226 babies with varying degrees of encephalopathy. (Robertson C, Finer NN. Term Infants With Hypoxic-Ischemic Encephalopathy: Outcome At 3.5 Years. Developmental Medicine & Child Neurology. 1985;27(4):473-84).

If we concentrate on the mild cases in that study, there were 79 who were evaluated for the presence of “handicap”; the definition of “handicap” used in that study included cerebral palsy of any severity, severe visual impairment or deafness, difficult to control epilepsy, or a score more than 3 SD below the population mean (100) on Stanford-Binet IQ tests. Using that definition, which picks out children whose condition has a significant impact on their lives, all the 79 mild encephalopathy cases were non-handicapped.

In contrast all the severe encephalopathy cases were either dead or “handicapped”, and the moderate cases had about 1/3 prevalence of “handicap”.

In those earlier studies Sarnat stage 1 refers to infants whose worst stage during their hospitalization was stage 1, children who deteriorate after the first few hours are not uncommon, and they would have been classified by their worst stage.

A recent publication from McGill describes short-term outcomes and MRI injury among infants who presented with mild encephalopathy, the study doesn’t include long-term follow-up, but there were a number of kids who had mild encephalopathy at presentation who had adverse clinical or MRI outcomes. (Gagne-Loranger M, et al. Newborns Referred for Therapeutic Hypothermia: Association between Initial Degree of Encephalopathy and Severity of Brain Injury (What About the Newborns with Mild Encephalopathy on Admission?). American journal of perinatology. 2016;33(2):195-202.) In this study there were 50 babies who had mild encephalopathy on admission, were not cooled and had an MRI. There were also 13 who had mild encephalopathy who were cooled, mostly because of aEEG abnormalities. Of the 50 not cooled babies, there were 20, (40%) who developed MRI abnormalities, including some very severe abnormalities, one of whom actually died of complications of HIE. Many of the babies who had MRI injuries had a worsening of their clinical status over the first couple of days of life, and none of those who were mild but were cooled because of aEEG abnormalities had MRI injuries.

Another fairly recent study from Parkland Memorial hospital described the short-term outcome of 89 babies who had acidosis at birth and were evaluated for the presence of encephalopathy. 69 of them had mild encephalopathy during the first 6 hours of life, (20 were normal) and of those 69 a substantial number had adverse short-term outcomes, which included later onset of seizures in 5, one of whom progressed to very severe findings and died of multi-organ failure. Others of them had feeding difficulties, persistent neurologic abnormalities at discharge, or abnormal MRI findings, for a total of 12 of the 69 with mild encephalopathy. The 20 with a normal exam were fine. Presumably, repeated encephalopathy exams would have found some progression in those who ended up with later abnormalities, and, of course, those with later seizures would automatically have been classified later on as having at least moderate encephalopathy.

A study just published on-line in Pediatrics has further data, (Murray DM, et al. Early EEG Grade and Outcome at 5 Years After Mild Neonatal Hypoxic Ischemic Encephalopathy. Pediatrics. 2016). This is a very well done prospective cohort study, all the babies were examined within  6 hours of birth by a single pediatric neurologist, all had 24 to 72 hour EEGs and there was a prospectively enrolled healthy full term comparison group who also had EEG (but only for 2 hours for obvious reasons). 53 babies with HIE were included, and 30 controls, the babies had been born before 2006 and were followed up to 5 years of age. The encephalopathy grade finally assigned was determined on the examination at 24 hours of age, which might not have been the same as the 6 hour grade, and may not be the same as the worst grade either.

The study shows a number of things, first of all the importance of having a comparison group. The comparison group had a Full Scale IQ of 117. So either the initial standardization of the test (WPPSI-III) no longer applies, or the controls were selectively of higher IQ than the Irish population, or (and this is my explanation) Irish babies are just smart. In comparison to the controls, the babies who were stage 1 HIE at 24 hours of age had lower IQ across all domains, and a mean FSIQ of very close to 100.


A Worse EEG grade at 6 hours was associated with poorer outcomes, EEGs all tended to improve or stay stable at 24 hours, and the grade of EEG abnormalities was more predictive at 24 hours than at 6.

So to summarize, infants who have stage 1, or mild, encephalopathy at 6 hours of age, when the decision is being made to cool them or not, do not have universally good outcomes. Many of them will have adverse outcomes, which can be predicted to some extent by: a later worsening of the clinical stage including later onset of seizures, or an abnormal aEEG or EEG, or persistence of stage 1 signs for more than 7 days (I didn’t talk about those data, they are older but appear reliable).

When comparing outcome studies in this group it is therefore important to know whether they are referring to the worst grade of HIE, or the grade at 6 hours, or at some other time point.

What to do about these babies? Starting hypothermia treatment later may actually have some effect, the 6 hour cut off that we use was rather arbitrary, and based on animal models with known timing and duration of the hypoxic ischemic insult and a fairly standard severe cerebral injury. Infants who have a mild encephalopathy and who start to deteriorate might therefore benefit from cooling started at 7, 8, 9 hours after insult.  But I presume there must be a limit, once those neurones are dead they aren’t coming back to life.  It seems that aEEG monitoring (and interpretation) for infants with mild encephalopathy is helpful in predicting which kids will progress to more severe short-term outcomes, according to, among other studies, the first study in this post, and in therefore deciding to proceed to cooling. If you can easily get a multi-channel EEG with expert interpretation as soon as the baby arrives, that is probably better than an aEEG.

The approach we have generally taken at Sainte Justine is that if the baby’s exam is close to qualifying, and we aren’t quite sure, then we usually cool them. Therapeutic hypothermia has a very favorable safety profile, sometimes PPHN gets worse, (and we once had to terminate cooling at 60 hours because the PPHN became uncontrollable and the only other option was ECMO).  Apart from that there are few complications unless the babies get too cold. Unfortunately babies with milder encephalopathy don’t like being cooled, they tend to seem uncomfortable, and often need some sedation which complicates their evaluation.

I think we need an RCT. I thought I would surprise you with that one! Infants with mild encephalopathy at 6 hours of age could randomized to be cooled or not, and stratified according to the aEEG patterns. Infants with mild encephalopathy before 6 hours could be reassessed at 6 hours to ensure they hadn’t progressed, and to avoid cooling babies who recover. They will then need to be followed quite a long time I think, the incidence of long-term problems is low enough that the sample size will have to be quite large.

I am still not sure that MRI findings are predictive enough of long-term outcomes that they could be used as a surrogate outcome (also known as a “biomarker”). I do think that if you found a positive effect of therapeutic hypothermia on MRI findings in mild HIE, that would probably justify more widespread cooling of mild encephalopathy; but if you found no MRI effect I think you would still need the clinical follow-up to be sure that there was no real benefit.

Posted in Neonatal Research | 2 Comments