When we counsel families about the potential outcomes for infants born profoundly preterm, I think we would all agree that we should be honest, transparent and truthful. Which includes, I would suggest, truthfulness about our own local results, as well as what is possibly available elsewhere.
In order to communicate such results we need to have a good idea of what those results are. I think it is entirely acceptable to say, for example “we don’t have much experience with babies of 22 weeks gestation, they don’t come along very often, and until recently we didn’t think that active care of such infants was a good idea. We have changed our approach and will do the best we can for your baby, whatever we choose together for her”. The rest of the conversation should follow the parents leads, while trying to determine what is important for them.
All of that, of course, requires that you have some idea about the likelihood of survival and of long term functional outcomes. Unfortunately many professionals in the chain of decision making are not well informed themselves. In a recent blog post I discussed an article which asked health professionals in the UK involved in neonatology to estimate survival and severe disability among survivors, and found major variability in estimates, generally estimates of survival were lower than a published calculator, and estimates of “severe disability” were higher.
Estimates of the respondents were compared to recent actual local data, including outborn babies. Overall, as in several previous studies, survival was underestimated, and “major disability” was dramatically over-estimated.
As this study again shows, the majority of survivors at every gestational age do not have major disability, whereas respondents think that the proportions are reversed. Obstetricians and nurses tended to be more pessimistic than neonatologists, another typical finding.
The authors compared their results to a survey from 2010, and found that respondents were more pessimistic than they were previously, when comparing estimates to contemporary results. Survival is improving, but it seems likely that mothers are receiving inaccurate, unnecessarily pessimistic information.
Even the New England Journal are getting in on the act (Lee CD, et al. Neonatal Resuscitation in 22-Week Pregnancies. N Engl J Med. 2022;386(4):391-3), I guess that someone talked to the editors about the practice variation in resuscitation of profoundly immature babies, and in response they have published this short vignette with 2 somewhat opposing views, Leif Nelin who promotes the idea that we should recommend universal active intervention, and Elizabeth Foglia who is in favour of recommending selective resuscitation.
I find it very interesting that there is not a 3rd author promoting an approach which still happens in many centres, i.e. recommending universal comfort care.
It is also interesting that there is no real disagreement on the facts, that without active intensive care mortality is 100%; that with intensive care some babies survive, and the majority of the survivors have good lives. The actual proportion of survivors is, of course, very variable, and it requires a commitment of both obstetrics and neonatology to work together to achieve the best results.
Dr Foglia says 2 things that require some reflection, she notes that “almost all extremely preterm infants require resuscitative interventions after birth to survive” which is sort of true, but depends on what you mean by “resuscitative interventions”, in most centres all such babies have endotracheal intubation shortly after birth, but further “resuscitative interventions” are uncommon. The second thing is “The current limit of viability is 22 weeks’ gestation.” That is stated as a verity, but it ignores 3 things, 1. we never know exactly what the GA is, except after IVF, so if you actively intervene for all 22 week GA babies, you will have intervened for some at 21 weeks. 2. If survival at 23 weeks can be as high as 60%, surely at 21 weeks and 6 days it would not suddenly drop to zero! 3. There are reported survivors who were thought to be <22 weeks.
Current guidelines do not often recommend antenatal steroids at 22 and 23 weeks, which is partly because of a lack of such infants in randomized controlled trials, but we are unlikely to have substantial numbers of mother in trials at those gestations for a while, if ever, so observational data are all we are likely to have. Rossi RM, et al. Association of Antenatal Corticosteroid Exposure and Infant Survival at 22 and 23 Weeks. Am J Perinatol. 2021(EFirst). This article, as one example, calculated the probability of survival at 22 and 23 weeks of GA, according to whether steroids were given prior to delivery. The data source they used had no information of timing of steroids, it was just a checkbox, yes or no. It probably includes, therefore, many babies with brief steroid exposure. Survival is only presented for babies who received active neonatal intensive care. The overall survival at 22 weeks, to one year of age, is shown below, divided by birth weight categories.
ANCS= antenatal corticosteroids
They don’t have the same sort of birth weight breakdown for the 23 week babies, but overall 1 year survival was 58% after antenatal steroids, and 48% without steroids. Relative risk 1.5 (95% compatibility intervals, 1.3-1.6). 62% of the 22 week deaths of babies who had antenatal steroids were before 7 days of age, as were 53% of the 23 weeks infants.
Currently all the data about such deliveries is consistent, ANS administration is associated with a major improvement in survival, the NNT is actually smaller than at any later GA. All the studies, unfortunately, suffer also from the same biases, which are sort of self-evident.
What is also consistent, is that centres with the best results, have a co-ordinated approach with obstetrics, and routinely give steroids as soon as the mothers are admitted.
As for my response to the NEJM article? I would perhaps phrase it a little differently, I think that active neonatal intensive care should be offered as an option to all mothers presenting with an increased risk of delivering at 22 to 24 weeks gestation, and that option should be presented as a reasonable choice which will be supported by the whole team, who will then do whatever they can to have the best possible outcome. When additional risk factors are present, such as growth restriction, imminent delivery without benefit of significant ANS exposure, then the discussion of the options must recognize those facts. When increased risk is very great, such as estimated weight <400g or florid chorioamnionitis, then it is vitally important to be realistic. It is also important to recognize that the decision to give steroids, as soon as possible, does not mandate active neonatal care, but will give the best chance for the baby if the later decision is indeed to proceed with intensive care. And that a decision for such care does not mandate a cesarean delivery, which should be considered a separate (obviously related) decision, which takes into account additional factors, including maternal age, risk factors etc.
This is a fascinating study from Columbus Ohio in mildly preterm babies when they reached term. (Richard C, et al. Randomized Trial to Increase Speech Sound Differentiation in Infants Born Preterm. J Pediatr. 2021) We know that prematurity and NICU care both have adverse impacts on language acquisition, and these authors wondered if there are ways we can mitigate those impacts. The babies were late preterms who were enrolled at term or slightly after (post-menstrual age of 36 to 52 weeks. They had a normal hearing screen using brainstem responses, and their families spoke only English. The investigators recorded female voices speaking American English, Mandarin Chinese, or French (they don’t say if it was Parisian French or Quebecois), with speech patterns such as those used for speaking to young infants, which they characterise as “a slower rate of speech, higher vocal pitch, simpler sentence structure, more variable prosody (rhythm, stress, and intonation of speech), and elongated vowels.” These are characteristics of infant-directed speech in many languages (and also of speech directed to our pets). The recordings also included lullabies in the relevant language, and the recordings were played to the babies for 15 minutes, twice a day for a couple of weeks.
They then tested the responses by playing synthesized syllables, consisting of a vowel and a consonant, from each of the languages, and measured auditory responses using temporal electrodes, obtaining the responses before and after the intervention.
Using this methodology, they were able to identify the brainstem responses to individual sounds, and also to see if pairs of sounds gave the same response, if responses differed it meant that there was differentiation between the syllable sounds.
Figure 3. Examples of Grand average tracings of ERP in response to different speech sounds for each language intervention group, pre- and post-exposure in T6. Only responses to representative speech sounds for the intervention language are shown, (eg, response to French sounds for the French-contingent intervention group). Examples of individual speech sounds are shown as dashed and dotted lines; absolute difference in mean amplitude between speech sounds is shown for each contrast (solid line). Differentiation of speech sounds is calculated as the absolute difference in mean amplitude between speech sound pairs in the 250 ms-400 ms poststimulus time window (grey bars). Increased speech sound differentiation occurs after interventions.
What they found was that exposure to the foreign language increased differentiation between pairs of sounds in that language, but not in the other language, and it had no negative effect on differentiation between pairs of English sounds.
As a background, you should realize that preterm infants have been shown to be able to differentiate syllables at 30 weeks gestation, and young infants can differentiate hundreds of speech sounds (phonemes), an ability which we gradually lose as we acquire a language.
This probably explains some of my difficulty in pronouncing French correctly, as I learned to speak French relatively late in life, I had already lost much of my ability to differentiate certain phonemes, an ability which is hard to relearn. To a native French speaker there is a world of difference between “tu” and “tout”, but it is much more difficult for others. Similarly, many native French speakers have great difficulties with my name as “…th” doesn’t exist in French, and their brains have difficulty in figuring out how to make that sound. Which leads to me frequently being called “Keet” or “Kees” by fellows in our program who are from France! Interestingly Quebec natives, who are generally brought up with a massive exposure to English during their infancy, have much less difficulty and can usually manage “Keith”. “th” both as a hard and a soft sound (‘they’ compared to ‘think”) are common phonemes in English, and Quebec infants probably have major exposure to them.
The research seems to give clues to ways that we might, in the future, enhance language development and perhaps overall executive functioning. In general terms, children from bilingual families have language development which mirrors that of unilingual families, if both languages are taken into account, and in the long term have superior language skills. There is evidence that a bilingual upbringing improves executive function, and there is evidence that that applies to preterm babies. At least one study has shown that preterm babies born to bilingual families have enhanced executive function skills copmared to the monolingual (Baralt M, Darcy Mahoney A. Bilingualism and the executive function advantage in preterm-born children. Cogn Dev. 2020;55).
I think it is now clear, even as research is ongoing, that reducing exposure to noxious sounds, and increasing exposure to human voice is important for enhancing future development of preterm NICU babies. It is important that we structure our NICU care to allow major exposure to human voices, reading to babies, talking to them using appropriate speech patterns, singing lullabies, and possibly talking to them in a second language. It is one thing we can encourage parents to do in order to be involved in their baby’s care and have a positive impact on their long term outcomes.
We are fortunate in neonatology to not have to worry about C difficile, partly because the name has been changed for this germ as well (now Clostridioides, rather than Clostridium), and it is one less name change to keep up with, but more seriously because it can be devastating, difficult to eliminate and frequently recurring. There are some parallels with NEC, however, including an association with prior antibiotic use, an association with gastric acid production inhibitors, a deranged intestinal microbiome preceding the condition, and a predilection for the most fragile patients. Probiotic prophylactic studies have generally shown efficacy, but one very large trial found no effect (I think because the control group rate was so low), and probiotic prophylaxis is not universal despite a favourable risk-benefit and cost-benefit profile.
There are also differences, including the clear efficacy of an intervention that has not been tried in preterm neonates (to my knowledge): fecal transplantation. And now a trial, just published in the NEJM (Feuerstadt P, et al. SER-109, an Oral Microbiome Therapy for Recurrent Clostridioides difficile Infection. N Engl J Med. 2022;386(3):220-9), of a probiotic preparation specifically designed to prevent recurrence using a mixture of spore bearing Firmicutes, was dramatically successful in reducing recurrence, from 40 to 12%, relative risk 0.32, compatibility intervals 0.18, 0.58. There were zero adverse effects. The role of bile acid metabolism was something I did not know about, but the Firmicutes metabolize primary to secondary bile acids, preventing the C difficile from having access to primary acids, a further analogy with the role of Bifidobacteria in metabolizing HMOs.
With the risk of giving too much press to one specific commercial strain of probiotic for the newborn, another recent publication regarding B longum susbsp infantis (Bajorek S, et al. B. infantis EVC001 Is Well-Tolerated and Improves Human Milk Oligosaccharide Utilization in Preterm Infants in the Neonatal Intensive Care Unit. Frontiers in Pediatrics. 2022;9) looked at stools of 15 VLBW infants before and 15 infants after introduction of probiotic administration in their NICU. None of the control group became colonized with the organism, compared to all of the intervention group, the more heavily colonized the baby was, the less Enterobacteria were present in their stools. The B infantis was very efficient at metabolizing oligosaccharides, leaving practically zero of the measured HMOs in the babies’ stools.
The abbreviations refer to commonly studied HMOs, 2′-fucosyllactose (2′-FL), 3-fucosyllactose (3′-FL), lacto-N-tetraose (LNT), Lacto-N-fucopentaose I (LNFP I) and 6′-Sialyllactose (6′-SL). I don’t know if the LNT is the same thing as the DSLNT (disialyllacto-N-tetraose) that I have previously discussed, but if so it is an extremely important HMO, and the ability to metabolize it may be important for probiotic efficacy.
We start probiotics along with the feeds (maternal or donor breast milk) in our extremely and very preterm babies, and I have often wondered whether it was effective to give probiotics with the feeds to those babies who were receiving antibiotics, or whether we should wait until the antibiotics have stopped. Although obviously not directly relevant to our preterm babies with our probiotics and our usual ampicillin/gentamicin treatment, this study suggests that maybe it is worth doing after all.
In addition to the short term results that I just posted about, the publication from the NICHD NRN describes the longer term outcomes of the babies of 22 to 26 weeks GA born between 2013-2016 who completed their 2 year follow up.
I was pleased to see a detailed summary of multiple different outcomes, despite the persistence of “NDI” as an outcome and its division into “severe NDI” defined by any of a Bayley-III cognitive composite score or motor composite score less than 70, GMFCS level 4 or 5, bilateral blindness, or severe hearing impairment, and “moderate NDI”, defined as any of a Bayley-III cognitive composite score or motor composite score of 70 to 84 or GMFCS level 2 or 3.
At the risk of boring my readers, I will quickly re-iterate why I am opposed to using this terminology, and these definitions, you can skip the next 2 paragraphs if you are already in agreement with me! An impairment is defined by the CDC as “an absence of or significant difference in a person’s body structure or function or mental functioning”, the WHO definition is “problems in body function or structure such as a significant deviation or loss”. A low score on a developmental screening test does not satisfy either of these similar definitions. Cerebral palsy of sufficient severity to impact function can be considered an impairment, hearing and visual losses also, but a low score on the Bayley Scales of Infant Development (BSID) is rather different. Most infants with low BSID scores will not have similarly low scores when intellectual abilities are measured later, whichever version of the BSID you are referring to, and whichever threshold you pick. As most infants categorized as “impaired” fall into that category because of low BSID scores, this is an important consideration. It would be better to call this Neurologic Impairment or Developmental Delay, to emphasize that, for a substantial proportion of infants with low BSID scores, it is indeed a delay in development, rather than a “cognitive impairment” which is a term often used to describe low BSID scores. Also, importantly, “severe NDI” is often used as shorthand for “handicap so severe that it makes neonatal intensive care futile” and is then often often conflated with death as an outcome which no-one would really want, and used for counselling parents.
Thankfully, this publication has a much more detailed description of outcomes of the babies between 22 and 26 weeks gestation, which is much more useful than just categorizing them into one of 3 arbitrary groups. In the supplement are these very informative figures for each of the BSID scores:
The red diamonds are the means, while the boxes show the median and upper and lower quartiles
The BSID scores are also described in this table, which give the scores on the 3 composites, with the means, medians and numbers below certain thresholds, by gestational age and overall:
The graphs and table show that the median BSID scores do not vary that much by GA, but there is a higher proportion with low and very low scores on each of the 3 scales at 22 and 23 weeks. This much richer summary of outcomes is much more informative that simplistic categorizations, and should be standard.
The GMFCS outcomes (gross motor functional classification scale, a way of grading cerebral palsy) from the same table, with the same order of columns, show substantially more cerebral palsy of moderate and severe degree as GA decreases.
Another thing that I really appreciated about this publication is the lower part of the table, which includes many things that are not often reported;
Columns from left to right, 22, 23, 24, 25, 26 weeks GA at birth, and overall 22-26 weeks
Hospitalisations, feeding competence, and medical equipment at home are all things that have a big impact on families, but are usually not reported. Most babies at all GA had some oral feeding at follow-up, but the most immature babies are much more likely to have limitations in feeding skills. Additionally mobility aids are more frequently required, which goes along with more frequently having a GMFCS indicating significant CP, and more frequently having BSID motor scores below thresholds. As for re-hospitalisation, more than 50% of them were for respiratory indications (data in the supplement) followed by surgeries, other infections, and then a variety of other indications, including 5% for growth and nutritional issues.
There is no information presented here about behavioural or emotional issues, even though they are of importance to families, but at 2 years of corrected age I can understand the difficulties in analysing them. It is impossible to tell from these presentations which infants had multiple problems; an infant with severe CP, blindness, major feeding problems and low BSID cognitive score, has a very different life course to 4 children with each of those problems individually. NDI categorizations don’t help with that either, as having any one of the supposed impairments qualifies the baby for the category, as would having multiple problems. In the future it would be helpful to have a summary of how many infants have 2 or 3 or more problems.
There is also no analysis of which risk factors are associated with more problems at follow up, probably there will be many more publications addressing those issues, but I can already predict that PVL and severe IVH are associated with worse motor function, that surgical NEC is associated with worse motor and cognitive scores, that late-onset sepsis is associated with poorer cognitive scores, and that BPD, especially if treated with steroids is associated with lower cognitive, language and motor scores. Preventing all of these complications will improve survival and will also improve long term outcomes.
Many of the results are compared with a previous cohort from the network, which had very similar enrolment criteria. In the previous cohort, survival to discharge of those born at 28 weeks was 94%, it was 90% at 27 weeks, 85% at 26 weeks gestation, and 77% at 25 weeks. The figures for the same GAs are now 98%, 98%, 90%, and 80%, therefore, even among these larger extremely preterm babies survival seems to still be trending upwards.
At 24 completed weeks GA 62% survived in the previous cohort compared to 70% in the new; almost all received active treatment in the recent cohort, that information isn’t given in the older study. At 23 weeks survival was 32% previously, and was 49% in the new article, with about 88% of the babies getting active treatment, among those 958 with active treatment survival to discharge (or one year of age if still hospitalised), was 56%. At 22 weeks there were 334 babies in the previous cohort with a 7% survival, which has improved now to 11% among all the babies delivered, but when restricting the analysis to only those 201 (of 550, 37%) who received active intervention, survival was 30%. A lot of the deaths of the 22 and 23 week infants were within the first 12 hours, (21% and 11% mortality at <12h of age). Many of the other complications and interventions are reported for those surviving >12h.
The network reports how many babies had limitations of life-sustaining interventions, and the percentage that this applied to was 36% of the babies born at 22 weeks (who survived >12h), progressively decreasing to about 2% at 28 weeks. Of the babies at 22 and 23 weeks gestation who survived >12h (n=159 and 856), 133 and 779 survived more than 3 days, 77 and 609 survived to 28 days, for a final numerical survival to discharge of 60 and 535. I recount these numbers just to point out that late death occurs among these infants, often after multiple complications, but the majority of deaths occur relatively early.
Complications of prematurity are also reported, including NEC, showing a small reduction from 10.3 to 8.9% between the 2 cohorts, and a relatively modest trend to increasing incidence with lower GA (5% at 28 weeks, 11% at 22, 24 and 25 weeks, 15% at 23 weeks). Severe intracranial hemorrhage does not seem to have changed overall at about 14%, despite the much higher survival at 22 and 23 weeks, who had an incidence of 38 and 36% respectively.
As many other cohorts have shown, there seems to be an increase in chronic lung disease, from 45 to 49% overall, with about 80% still on oxygen at 36 weeks for babies at the lowest 3 weeks of GA. As an exercise I calculated “death or BPD”, at 22 weeks 97% had this outcome, and at 23 weeks it was 95%.
Cystic PVL was unchanged from the previous publication and did not differ across GAs,at around 4%.
During the recruitment of this cohort, there were a few things changing in neonatology, some of which should lead to further improvements in these figures:
Delayed cord clamping was being introduced during this period, and in this cohort only about 40% had delayed clamping. This intervention has been shown to improve survival, so future cohorts should benefit from this. In this cohort some received cord milking, which I think the data shows should not be used as a replacement for delayed clamping.
Antenatal steroids were only given to 30% of the babies who delivered at 22 weeks, and 81% of the 23 weekers, compared to about 90% to the more mature babies. Proactive care co-ordinated with the obstetric team can have major impacts on survival of the most profoundly immature. Although such babies were not included in the majority of RCTs of antenatal steroids, all the observational data show a survival benefit of steroids at 22 and 23 weeks. They can be given as soon as a mother presents with threatened extremely preterm delivery, in order to give time for them to have an effect, even if a decision regarding active intensive care has not yet been taken.
Probiotics were given to a minority of the babies, only 10%, if you read this blog you will know my opinion on this, the network meta-analysis showed a substantial benefit of a mixture containing Bifidobacterium longum susbsp. infantis on the frequency of NEC.
There were many babies who did not receive human milk; I think that donor milk is now universally available, which may not have been the case at the beginning of this cohort, but should also have an impact on NEC. Of all the interventions that we must ensure are applied universally to very preterm infants, human breast milk feeding should be top of the list.
Late onset sepsis is depressingly very common in the most immature babies, most centres have on-going quality control initiatives to reduce nosocomial sepsis, and although important, they have been of limited impact in the most extremely preterm infant. In the Canadian Neonatal Network, quality improvement initiatives using the EPIQ framework have led to a progressive decrease in late-onset sepsis for babies of >25 weeks. If you compare the following 2 graphs from 2013 and 2020, there is a substantial reduction in late-onset culture-positive sepsis between the 2 years, until you reach the <25 weeks stratum, where the incidence seems to have gone up. These data are among infants who survived more than 2 days, and are the proportion who experienced at least one culture-positive sepsis, about 25% of whom had more than one sepsis episode. Part of this worsening may be due to the increased survival of infants at 22, 23 and 24 weeks gestation, as there are many more patient days at risk in 2020 than there were in 2013.
2013 national results, CNN annual report2020 national results, CNN annual report
These results point out the extreme importance of further research in ways to reduce sepsis, by improving skin and intestinal barrier function (some studies showing a reduction in LOS with probiotics, and meta-analysis usually confirming a benefit), and supporting and improving immune function. An interesting discussion from a Japanese neonatologist (Isayama T. The clinical management and outcomes of extremely preterm infants in Japan: past, present, and future. Transl Pediatr. 2019;8(3):199-211) points out that gloving, masking and gowning are often universally applied during routine care in Japan, where the incidence of late-onset sepsis is very much lower than in North America and Europe. It makes me wonder whether that approach would be worth instituting, especially for the highest risk babies, an whether the recent extra restrictions imposed by the pandemic might have had an effect on sepsis. One advantage of universal extra precautions would be to inhibit people from touching and disturbing the babies. Previous data have suggested that universal gloving, applied in an NICU because of a hospital-wide policy during RSV season, reduced late-onset sepsis. Randomized controlled trials have confirmed this benefit.
And of course, decreasing lung injury: although cause of death is not reported in this cohort, being very difficult to ascertain in a large database, many infants who die late have severe respiratory failure. If we can improve lung injury we may well improve survival, and certainly will decrease adverse long-term pulmonary outcomes which have a serious impact on health-related quality of life. What are the potential hopeful interventions for reductions in lung injury? Prophylactic budesonide, given with surfactant, is one potential that is being studied in several multicentre trials. Finally figuring out when postnatal steroids should be given, which molecule, in what dose, for how long, would be great. Late surfactant therapy in babies still intubated who are starting to develop chronic lung disease has not been adequately studied, but seemed to improve long term lung health in the French study.
Unfortunately, for many interventions, the most immature babies, who may not have the same response to more mature preterm infants, have been excluded from trials. Hopefully, in the future there will be no minimum GA or birth weight for eligibility; surely any baby for whom intensive care is instituted should be eligible for research to try and improve outcomes for future similar babies.
In the next post I will discuss the longer term outcomes from this publication.
Rao S, et al. Probiotic supplementation in neonates with congenital gastrointestinal surgical conditions: a pilot randomised controlled trial. Pediatr Res. 2022. When we started using probiotics there had been a couple of case reports of probiotic associated sepsis in infants with congenital GI anomalies, specifically omphalocoele, so we decided to not include them in our probiotic protocol. I was never too sure that was the right thing to do, especially as infants with gastroschisis in particular are at high risk of NEC. The group from Perth, Western Australia, randomized 61 babies to a Bifidobacterium mixture (B. breve, B. longum longum, and B longum infantis) or placebo and followed their intestinal microbiome development weekly until discharge, they found that there was a progressive change in the microbiome due to supplementation, with less dysbiosis, specifically a lower abundance of potential pathogens (Clostridiaceae, Enterobacteriaceae, Enterococcaceae, Pseudomonadaceae, Staphylococcaceae, Streptococcaceae &Yersineaceae) and much more Bifidobacteria. In a pilot study of this size one wouldn’t necessarily expect to see a clinical impact, although growth parameters were improved with probiotics, in particular head growth; or rather, the postnatal growth restriction, was less severe in the probiotic growth.
Plummer EL, et al. The effect of probiotic supplementation on the gut microbiota of preterm infants. J Med Microbiol. 2021;70(8). This study analysed the intestinal microbiomes of infants randomized in the ProPrems trial, which was a large multicentre trial of a mixture of B longum infantis, B animalis lactis and Streptococcus thermophilus vs placebo. 99% of the babies in the probiotic group had the Bifidobacterial DNA in their stools during treatment, but in only about 90% of the fecal samples, about 20% to 35% of the controls also had the bacteria in their stools at times. The relative abundance of the B infantis was quite variable, suggesting we need to find ways to make colonization more effective.
Hui Y, et al. The effect of early probiotic exposure on the preterm infant gut microbiome development. Gut Microbes. 2021;13(1):1951113. This is a study of microbiome development in 2 cohorts of preterms, before and after routine probiotic use (a mixture of Lacticaseibacillus rhamnosus GG, the bacterium formerly known as Prince Lactobacillus, and Bifidobacterium animalis ssp. lactis) in Copenhagen. They showed less impact on the progression of microbiome changes than some other studies, associated with a variable abundance of the probiotic organisms, but did show an impact leading to a reduction in Klebsiella concentrations in the stools.
Fleming P, et al. Bifidobacterium breve BBG-001 and intestinal barrier function in preterm babies: Exploratory Studies from the PiPS Trial. Pediatr Res. 2021;89(7):1818-24. This is an analysis of microbiome impacts and intestinal functional impacts of the B breve used in the PiPs trial. In addition to the usual 16sRNA, they also cultured the stools, and found they were able to grow Bifidobacteria, identified by pcr as B. breve, from most of the probiotic babies, as well as some controls, but found that babies colonised with B breve actually had more potentially pathogenic Enterobacteriacae from 3 weeks onwards. They did not see a benefit on intestinal barrier function of the probiotic.
Overall, I think these recent studies imply that the best probiotic to have an impact on the preterm infant microbiome, and, as a result on NEC, is Bifidobacterium Longum subsp. Infantis. Other data from network meta-analysis that I have discussed previously, is also consistent with this, and also consistent with a benefit of a mixture of Bifidobacteria (most mixtures include B infantis). Ways of ensuring good colonisation, such as the addition of prebiotics, especially Human Milk Oligosaccharides, and provision of human milk will probably help to ensure the greatest benefit.
I haven’t, surprisingly, posted about probiotics recently, but there are some new findings in the literature that warrant discussion.
Granger C, et al. Necrotising enterocolitis, late-onset sepsis and mortality after routine probiotic introduction in the UK. Arch Dis Child Fetal Neonatal Ed. 2021. The majority of pre- post- studies of introduction of probiotics that have been published have shown a significant reduction in NEC. Of course, there could well be a huge publication bias in such publications. I am pleased therefore, that this study was published, even though it showed little impact on NEC, overall. Especially as it is from a centre with a huge interest in intestinal microbiome development in the preterm, and in NEC, that is Newcastle upon Tyne, UK. They showed a small reduction from 10.6% of babies <32 weeks to 9.2% after routine introduction of probiotics, intially Infloran (Lactobacillus acidophilus and Bifidobacterium Bifidum) then Labinic (same species plus B longum ssp infantis)in the last 30% of the treated cohort. Overall there was no major impact on their outcomes, but subgroup analysis suggested a reduction in NEC in the more mature babies >28 weeks, and a reduction in late-onset sepsis in the less mature <28 weeks. Overall mortality, and mortality associated with those complications, were not affected.
Tobias J, et al. Bifidobacterium infantis EVC001 Administration Is Associated With a Significant Reduction In Incidence of Necrotizing Enterocolitis In Very Low Birth Weight Infants. J Pediatr. 2022. In this centre, in contrast, there was a dramatic reduction in NEC when they started routine supplementation with B. infantis. The strain they used, noted in the title, is a commercially produced strain produced by a company which sponsored this study, and of which one of the authors is an unpaid consultant. In this centre in Oregon, they showed a reduction in NEC from 11% to 2.6% of VLBW infants after introduction of the probiotic. Mortality related to NEC was also reduced, from 2.7% to 0. The impact applied to the ELBW subgroup also, 19.2% NEC incidence reduced to 5.3%. The relative risk of NEC after probiotic introduction was 0.27, compatibility intervals (0.09-0.61).
Another study using the same strain is Nguyen M, et al. Impact of Probiotic B. infantis EVC001 Feeding in Premature Infants on the Gut Microbiome, Nosocomially Acquired Antibiotic Resistance, and Enteric Inflammation. Front Pediatr. 2021;9:618009. Of note, many of the authors were employees of the company which makes the probiotic preparation, and they also funded the study. In this project VLBW babies <32 wks were given probiotics, bigger preterm babies were not. They also differed slightly in diet, as they all predominantly received human milk, but formula was only given to the larger babies that needed a supplement, rather than donor milk given to the smaller ones. There are several very pretty figures in the publication, some parts of which I actually understand! The following for example, graphic B, I think, shows the overall percentage of all the bugs they found in all the stools, that were in each of the 3 groups noted by the colours, I couldn’t tell you exactly what they are percentages of, as all the other OTUs are missing from the graph, but it is designed to convince you that there are a lot more Bifidobacteria around.
As for panels C and D, they are from the 2 GA groups, C from the more immature babies who received the probiotic, and D from the bigger babies. They also looked at whether the stool samples from the babies had bacteria capable of metabolising HMOs, and I will have to take their word for it, but it sure looks like the babies fed the probiotic had more ability to metabolise HMOs.
The 2 centres involved had a different pattern of antibiotic resistance genes, which the babies acquired throughout their stays in the NICU. As a non-specialist, I am not sure how well known this is, it certainly was something I was not aware of, but it is not a surprise, that different NICUs have different antibiotic resistance gene populations. Different NICUs tend to have differing bacteria causing their sepsis, and intestinal colonisation during NICU stays. They also showed that babies transferred from one hospital to the quaternary centre acquired the receiving centre’s antibiotic resistance genes after they arrived. The B infantis reduced the abundance of those genes compared to more mature, untreated controls. There was also, finally, a reduction in pro-inflammatory biomarkers among the immature, probiotic receiving babies. They state that their data show that gut microbiome composition can be altered by feeding probiotics. But the data don’t show that. As there are no randomized controls, all they can say is that the 2 groups are different, and that there is an association with probiotic administration. Which is my major criticism of all these EVC001 trials, why haven’t they performed an RCT powered to detect a reduction in NEC? There are still many centres in the USA who don’t use probiotics, partly because of the lack of an FDA licensed preparation containing, probably, the most important species of probiotic organism for the newborn, which is extremely well standardised, with the quality control of a medical product, like this one.
They might be able to convince the FDA, for that latter requirement, to give them a license for preterm babies, for the prevention of NEC, if their trial proved the value of Bifidobacterium longum susp infantis, strain EVC001 for the prevention of this devastating condition in preterm infants.
As many of you will know, I have advocated for many years for evaluation and management of very preterm babies based on their tissue perfusion rather than the blood pressure. There is little or no correlation between BP and perfusion, and many of the commonest interventions, in particular dopamine, may increase BP almost exclusively by vasoconstriction, thereby worsening tissue perfusion. Such an approach, however, suffers from the difficulty in evaluating tissue perfusion in the preterm infant, clinical signs are limited, and lab tests may be misleading. Echocardiography is relatively objective, but difficult, has some inter-rater variability, and multiple repeated studies are often needed during management, leading to major disturbance of fragile babies.
A reliable continuous method with minimal invasiveness would be a major step forward. The perfusion index of some pulse oximeters excited me a few years ago, and a study that my group performed showed a useful correlation of perfusion index in limited circumstances, at 6 hours of age among babies with a low SVC flow.
A group from Saitama in Japan have been studying laser doppler skin blood flow for a few years. They have described the normal transitional changes in flow, and noted that low flow may precede later intraventricular haemorrhage. In contrast to my statement above, they looked at the impact of dopamine on skin blood flow, and showed that dopamine appeared to increase both blood pressure and skin blood flow, at least blood flow in the lower limbs (increased by 5%), blood flow to the skin and subcutaneous tissue on the forehead was not affected, despite a 10% increase in mean blood pressure; I don’t know of any other study that shows an increase in systemic perfusion with dopamine in compromised preterm babies, but even I must admit that the currently available data are poor. If mean BP increased by 10%, but perfusion locally increased by between 0 and 5% that is some evidence of vasoconstriction, but perhaps locally in some vascular beds there may be a sufficient increase in perfusion pressure to improve perfusion.
The big question with any such method, is whether using the method can improve management to the extent that outcome are better. We know that, statistically, babies with lower blood pressure have worse outcomes and that babies who receive treatment have worse outcomes than those who do not; which means that treatment is either a risk factor itself, or is a marker of increased risk. We tried to differentiate between those possibilities in the HIP trial, but were unable to enrol enough babies to find a definitive answer.
The 440 VLBW infants were randomized to one of 2 contrasting protocols within 6 hours of birth, if they had no IVH on screening head ultrasound. Not all the infants had invasive blood pressure monitoring which is a reasonable pragmatic choice (and is, in fact, what I wanted to do for the HIP trial), but immediately introduces uncertainty, as non-invasive blood pressures are somewhat unreliable, especially in the smallest babies, and especially if the baby is hypotensive. All babies had skin blood flow measured from one foot, and were treated by the assigned protocol if the BP was below the gestational age in weeks, or if the BF (blood flow) was below 14 mL/min in the first 24 hours or 17 mL/min thereafter; depending on group. If they passed their treatment threshold then a functional echocardiogram was performed (in addition to a routine daily echo) and treatment given according to the schema above. The primary outcome was any intraventricular haemorrhage. They also evaluated how often the treatment was successful in improving the randomized parameter to above threshold within 3 hours.
The trial was a single centre study that took 7 years to complete, randomizing about 2/3 of eligible admissions. 37% of the BF group and 42% of the BP group passed their treatment threshold and had intervention. In the BF group they almost all had dobutamine, and nearly half of those also had dopamine. In the BP group about half of those with intervention received dopamine, and a similar proportion received dobutamine. In both groups 12% had a volume bolus and 6% had hydrocortisone.
I would question some of the reasoning that went into constructing these algorithms, but overall, I think they are not unreasonable. On the other hand, I find it a bit surprising that one in five VLBW infants had low blood flow with normal contractility with decreased preload indices and didn’t respond to a fluid bolus (as that is the only way a BF baby could get dopamine added to their dobutamine); which makes me wonder how well the algorithms were actually followed. Indeed I find it a bit surprising that such a high proportion of VLBW infants were considered to need cardiovascular support, they were enrolled early in life and for 40% of the babies to need some cardiovascular intervention in the first 72 hours seems aggressive. They do however have a historically low proportion of babies with IVH, in the CNN about 30% of VLBW babies have at least a grade 1 IVH, small subependymal haemorrhages being included in their outcome of IVH. Overall the outcomes are the typically very good outcomes of Japanese centres with a remarkably low mortality, with 1 death out of 440 VLBW babies within the first 7 days, and only 7 deaths prior to discharge, very low NEC and low late-onset sepsis are also evident.
The sample size was calculated based on that previous incidence of IVH of 13%, with a power to detect a reduction to 6%. I presume, though it is isn’t mentioned, that the previous IVH rate was in the context of BP-directed management of the infants. They actually showed an 11% rate of IVH in the BP group, and a 7% incidence in the BF babies. The 95% compatibility limits for the difference in IVH rate (not given in the manuscript) are a 9.4% absolute risk decrease in IVH and a 1.2% increase.
The study included some babies at relatively low risk, so the infants of 1001-1500 g birth weight only had 2% vs 3% IVH rate. It would be just about impossible to design a trial to find a reduction in an adverse outcome from 2%. Among the ELBW babies there were 18% IVH in the BP babies, and 10% with BF targeting, which give 95% compatibility limits of -17% and +0.5% absolute risk difference.
The intervention was not always successful at increasing the target parameter within 3 hours of initiating treatment, and a very interesting post-hoc analysis is presented, which shows that the BP and BF group babies who never received intervention had very low rates of IVH, (3% vs 6%) but those who had a successful increase in their BF did better than those who had a failed application of the protocol (i.e. BF stayed low at 3 hours of intervention) whereas the low BP babies who continued to have a BP below GA at 3 hours of age did better than those in whom BP increased.
This obviously has to be interpreted with a great deal of care, but does suggest several possible interpretations.
This study helps in calculating sample sizes for future trials, 11 of the 39 IVHs occurred in babies who had no cardiovascular intervention. I don’t have any way of calculating how many of those babies had both normal BF and normal BP throughout the trial, but a baby who stayed above both thresholds throughout 72 would have had no change in management. If we guess that about 40-50% of babies will have no cardiovascular intervention regardless of which group they are in, and perhaps 25% of IVH occurs in those babies, then that impacts the numbers of babies required for a trial to show that intervention algorithm reduces IVH frequency. It is complicated, of course, because some babies with low BP will have normal flow, and vice versa.
I think this trial is clearly not enough evidence for a universal change in practice, but it is strong evidence that we need to do more investigations of this way of managing cardiovascular support in the newborn. Refinement of the protocols, comparison with other indicators of perfusion (perhaps cerebral NIRS, perhaps perfusion index) and a larger sample size, or perhaps concentrating on the babies at highest risk, are needed to answer the vitally important questions posed by this study.
That is the question addressed by this follow-up of babies enrolled in the STOP-BPD study. As you may remember, that study used a 22 day tapering course of hydrocortisone, starting at 5 mg/kg/d for 7 days, and had the beloved primary outcome “death or arbitrary dichotomised definition of lung injury”. The trial showed no real difference in the primary outcome 71% HC, 74% control, but mortality diverged starting at about 7 days after enrolment, and was “statistically significantly” different at 36 weeks PMA (16% vs 24%), but not quite “significant” by discharge (20% vs 28%).
As I mentioned previously, the trial included very few babies under 24 weeks, as they were not receiving active care in Holland, where the study was mostly performed, and active care was rare in the Flanders region of Belgium which accounted for most of the non-Dutch centres, Liège and Charleroi were also involved, which are in Wallonia, but I don’t know what their approach was or if they enrolled any babies <24 weeks. As I have also mentioned, and has been shown many times, centres which don’t intervene at <24 weeks, or at <25 weeks, have poor outcomes at 24, or 25, weeks compared to centres which are active at lower gestational ages. Survival for the study participants was much lower than I would expect in my centre, even taking into account the selection of a higher-risk group, ventilator dependent on oxygen at 7 to 14 days of age.
Another problem with this study, especially with regards to long-term outcomes, is the frequent treatment of control babies with steroids (principally hydrocortisone) either after the study drug period, according to protocol, if they were thought to be at high risk of dying, or outside of the protocol. 108 of the original 190 control babies received hydrocortisone during their hospitalisation.
By the 2 year follow up there were a small number of extra deaths in each group, so the mortality was now 22% (HC) vs 30% (control); 95% confidence limits for absolute difference in percentage mortality are -17% to +1%.
There were no substantial differences in any measured developmental or neurological outcome. A dichotomised “NDI”, largely driven, as usual, by scores on the Bayley Scales of Infant Development (version III), was found in 44% of hydrocortisone, vs 47% of control survivors. I presume that the total dose of HC given to the HC babies was probably higher than the total dose given to controls, so this does give some reassurance that development, including motor and cognitive development as measured by Bayley, was not adversely affected. Cerebral Palsy, blindness and deafness were all a little less common among the HC babies.
Despite the limitations, this trial does suggest that long term development is not very seriously affected by the intervention tested, a total of 72.5 mg/kg of hydrocortisone, among infants of 24 weeks and more who are considered to be at high risk of BPD at 7 to 14 days of life, at least in comparison to relatively liberal use of rescue hydrocortisone if there seems a higher risk of death.
There is a modest difference in mortality favouring the HC group, despite rescue HC among controls. Can a similar study be performed in centres with active treatment at 22 and 23 weeks? Centres which will likely already have a lower mortality than this study’s intervention group, but which probably already use some steroids in babies at high risk? (In the CNN all the centres with very immature babies have some use of steroids for BPD, but it is very variable). It would be difficult to perform such a study, and some rescue use of steroids will have to be allowed in order to make the study feasible, but I think it is essential in order to answer the questions we still have about when to give steroids for lung disease, which dose, of which molecule, for how long.
Neonatal Research 