Caffeine is good for the preterm brain; might more caffeine be even better?

Posted on 24 September 2025 by Keith Barrington

One of the pivotal RCTs in neonatology was the CAP study (Schmidt B, et al. Long-term effects of caffeine therapy for apnea of prematurity. N Engl J Med. 2007;357(19):1893–902). We performed that study because there was no data on the long term impacts of caffeine, and there was a worry that blocking adenosine receptors in babies having multiple hypoxic episodes might be a bad idea. Adenosine is an inhibitory neurotransmitter that is produced during hypoxia, and decreases the brain metabolic rate to protect against hypoxic damage. So giving caffeine to babies having a lot of apnoeas could potentially have been a bad idea.

As it turned out, caffeine, given for a few weeks in the neonatal period, to babies <1250g birth weight who, in the first 10 days of life were thought to have an indication for caffeine by the attending physician, had a lasting positive impact, with some fine motor benefits even out to 11 years of age (Murner-Lavanchy IM, et al. Neurobehavioral Outcomes 11 Years After Neonatal Caffeine Therapy for Apnea of Prematurity. Pediatrics. 2018;141(5)). Why is caffeine so beneficial? It could be because of a reduction in apnoea, and in the consequent intermittent hypoxia, it also appears to have effects on cerebral oxidative injury and on apoptosis. There is one article, for example, from a study in mice, which showed that caffeine reduced hypoxia-induced white matter injury (Back SA, et al. Protective effects of caffeine on chronic hypoxia-induced perinatal white matter injury. Ann Neurol. 2006;60(6):696–705).

The dose of caffeine that we gave was based on the data available at the time, it appeared to have a wide margin of safety, so the standard dose that we settled on, a 20 mg/kg load of caffeine citrate, followed by 5 mg/kg/d, could be increased to a maximum of 10 mg/kg/d if thought to be clinically indicated. I published an abstract, and presented orally, at the PAS meeting in 2010, which showed that infants who got a higher dose, compared to controls who had a higher “dose of placebo”, had the same benefits on long term outcome that we saw among the overall group. To my shame, I never followed up and wrote up the full publication. But, as I was writing this post I dug out the powerpoint presentation that I gave in 2010 of this secondary analysis. I thought I would give you all a treat and show some of the results. We had 1583 infants with data for these analyses, and the medians for caffeine exposures are shown below:

An example of the results is given below.

This shows the percentage of infants in each of the two groups who had a low Bayley (version2) MDI score, <85, at 18 months corrected age, divided by duration of treatment. Infants who had treatment longer than the median of 45 days, with either caffeine or placebo, were more likely to have low scores than those with shorter treatment. It also shows that the difference between caffeine and placebo groups was the same, regardless of duration of study drug administration. We did the same kind of analysis for maximum daily dose received, total accumulated dose, average daily dose, and the PMA at which caffeine was stopped. None of the analyses showed any impact of caffeine exposure on the advantages of the caffeine group. The ORs in the figure above are the Odds Ratios for “cognitive delay” between caffeine and placebo in the 2 subgroups.

For the much smaller numbers with cerebral palsy, the results were different: longer duration of caffeine use, and higher total dose received were associated with a much greater difference between caffeine and placebo groups.

When we looked at average daily dose, or maximum dose received there was no association with the impact of caffeine on CP.

In other words, infants in the placebo group were more likely to develop CP (GMFCS grade 1 or worse) if they received study drug for longer than the median, but those who received caffeine were not. This made us wonder if the impact of caffeine on motor function was by different mechanisms to the impact on cognitive development. The overall primary outcome (which was “death or NDI”) was not associated with any of the metrics of caffeine exposure; the outcome, as usual, was driven by lower Bayley scores, so the impact of duration of therapy on CP was not noticeable in the composite primary outcome.

I must acknowledge the collaborators in the CAP trial here, the co-authors of that abstract were myself, Robin Roberts, Barbara Schmidt, Elizabeth Asztalos, Aida Bairam, Arne Ohlsson, Koravangattu Sankaran, and Alfonso Solimano, as well as the other CAP investigators; and of course the PI and driving force behind the trial was Barbara Schmidt. The title of the abstract was : The Caffeine for Apnea of Prematurity (CAP) trial: analysis of dose effects.

As I mentioned above, the enrolment criteria included use of caffeine to aid extubation, to prevent apnoea, or to treat apnoea. Some infants therefore received caffeine early as prophylaxis against apnoeic spells, others later after apnoea had become evident. Peter Davis led the group to publish a subgroup analysis, (Davis PG, et al. Caffeine for Apnea of Prematurity trial: benefits may vary in subgroups. J Pediatr. 2010;156(3):382–7), showing that the subgroup who received caffeine for prevention of apnoea had fewer benefits, in terms of the long term advantages, than those who received it for treatment of apnoea or to assist in extubation. In contrast, the babies who received caffeine earlier (less than the median of 3 days of age) had more advantages than those who received it later (>=3 days), they had the biggest decrease in duration of oxygen treatment, and therefore in having a diagnosis of BPD, and more long term advantages.

The CAP study was performed in the days before we were so aggressive with non-invasive support, so large numbers of the early treated babies were intubated at the time they were enrolled. There is therefore some overlap with another secondary analysis in that paper, which was the impact of caffeine according to the respiratory support at the time of randomization. Intubated and CPAP babies had more benefit from caffeine than those not on any support.

As I was preparing this post I read a Systematic Review and Meta-analysis which purported to be an SR of prophylactic caffeine, compared to control groups. Miao Y, et al. Effect of prophylactic caffeine in the treatment of apnea in very low birth weight infants: a meta-analysis. J Matern Fetal Neonatal Med. 2023;36(1). It was, however, seriously flawed, and I won’t provide a link. The authors had included the CAP babies as if they had all received prophylactic treatment, and had also included, a second time, the babies in the CAP prophylactic subgroup from the Davis article, who were therefore double counted. That wasn’t all, 4 of the references in the reference list, supposedly to articles included in the SR, were not to the right articles, which meant that 4 of those that were actually included had no reference, and no way to find them, including one study (Ke H 2018) which apparently had over 1000 VLBW babies. A failure of peer review, I’m afraid.

A good quality SR, in comparison, is this analysis of caffeine and dose effects, on long term outcomes. Oliphant EA, et al. Caffeine for apnea and prevention of neurodevelopmental impairment in preterm infants: systematic review and meta-analysis. J Perinatol. 2024;44(6):785–801. The results of analyses for BPD, PDA, and survival, for caffeine compared to placebo, were almost entirely dependent on the CAP trial (weight of over 95%). When they analysed the dose comparison trials, higher doses led to less apnoea and less BPD.

Another reasonably good SR was a comparison of caffeine use by timing, early (<3 days) vs late (3 days or later). Karlinski Vizentin V, et al. Early versus Late Caffeine Therapy Administration in Preterm Neonates: An Updated Systematic Review and Meta-Analysis. Neonatology. 2024;121(1):7–16.It only included 2 small RCTs, each with about 90 subjects, so most of the data are from the 9 observational studies. All of the outcomes were better with early treatment, except for mortality. As the authors point out, mortality may be higher in the early treatment group because you have to survive at least 3 days to be in the late treatment group, i.e. there was a survival bias.

One study that will help a lot with the use of caffeine is the ICAF trial, which was presented at PAS in 2024, but has still not published their results. The intervention was restarting caffeine or placebo at 36 weeks, after the clinically used caffeine was stopped, until 42 weeks 6 days PMA. Outcomes were the number of Intermittent Hypoxia episodes, inflammatory markers, and MRI findings at about 45 weeks. It seems to be taking a long time to publish, as enrolment had already finished in June 2023. The initial sample size was planned to be 220, according to the registration documents, but only 160 were reported in the abstract. As it was an abstract there aren’t all the details about why the sample size was not achieved. Nevertheless, I think the benefits of continuing caffeine to 43 weeks that were reported are very interesting, and warrant a consideration of our usual approach. We could do with some more information, though, before prolonging caffeine for all babies <30 weeks. Are there risks of stopping caffeine at home after discharge? Are there benefits other than those reported, which have a clinical impact? The outcomes reported in the abstract are very interesting (go here for a review) there was much less intermittent hypoxia, but whether that translated into a useful clinical benefit is uncertain, what was an important benefit, however, was an 11 day shorter length of hospital stay. If that is due to the intervention, which seems probable, even though it was an “exploratory” analysis, then that is a potential major advantage of continuing caffeine much longer than we usually do.

The reason for ruminating about caffeine use is that we still don’t really know the optimal dose (perhaps higher doses would be better) the optimal timing (earlier may well be better, but the data are very weak) the optimal duration (longer might be better, ICAF will help, and hopefully there is along term outcome plan for the ICAF babies).

A new observational study (Ostrem BEL, et al. Cumulative caffeine exposure predicts neurodevelopmental outcomes in premature infants. Pediatr Res. 2025) examined the total dose of caffeine received in babies at UCSF. They calculated the average daily caffeine dose of a cohort of infants <32 weeks who had TEA brain MRI. Average dose was used to convert the babies into 3 tertiles of dose received, infants had Bayley version3 scoring at 30 months corrected age.

The infants receive a standard bolus dose of 20 mg/kg, the median maintenance dose was 7.6 mg/kg/d (IQR 6.3- 8.7) which is higher than the starting CAP trial dose, of 5 mg/kg. So, many of these babies had higher doses, with a range therefore between about 5 and 10 mg/kg. The 3 tertiles included 23 babies in each group. They also divided the total caffeine exposure by the number of days between birth and the due date, to give what they called average daily caffeine exposure. This averaged about 3.3, as it included a variable number of days that the infants received no caffeine.

As you can see from those violin plots, infants who had more caffeine had, in general, higher scores on motor and language domains, and slightly higher scores in the cognitive domain. The data were corrected for Gestational Age and duration of oxygen therapy. There was also, interestingly, very poor correlation between MRI findings and Bayley scores, which I will come back to in a future post.

These new data are again suggestive of a developmental benefit of caffeine in the long term, with infants receiving more caffeine having higher scores, after correction for potential confounders.

The major concern about higher standard doses that I have comes from the 1 trial which showed an adverse impact of very high doses. McPherson C, et al. A pilot randomized trial of high-dose caffeine therapy in preterm infants. Pediatr Res. 2015;78(2):198–204. In that trial the loading dose was a whopping 80 mg/kg, over a 36 hour period, compared to 30 mg/kg over 36 hours in the comparison group. Both groups received maintenance of 10 mg/kg/d. The high dose in that study led to more cerebellar hemorrhages on MRI, and had more abnormal signs at term, although longer follow up did not show a disadvantage of the high-dose group.

You may recall that the first neonatal comparison in the PLATIPUS platform is between 3 different caffeine dosing groups, with low dose being the standard of 20 mg/kg load followed by 10 mg/kg/d, medium dose being 30 mg/kg followed by 15 per day, and high dose being 40 mg/kg followed by 20 mg/kg/d. Treatment is up to 36 weeks followed by open label treatment according to the teams preference. The hierarchical composite outcome of that study should capture both the potential hazards suggested by the McPherson study, and the possible advantages of higher doses. Of course the optimal duration of therapy will remain uncertain until further studies similar to ICAF are performed.

Posted in Neonatal Research | Tagged , , , , | Leave a comment

Unethical research practice, fraud and abuse of trust.

Posted on 2 June 2025 by Keith Barrington

One of the worst kinds of unethical research practice is to fail to publish results after a prospective study.

Parents consent to research for altruistic motives, in the belief that their baby’s participation will help the care of other, future, babies. Failing to carry through and publish, or at the very least, make results publicly available on the registration website, is an abuse of that consent. It is fraudulent to ask for consent for a prospective study and then to hide the results because they are not what you wanted to find.

This is one of the major reasons behind the mandatory registration of prospective research on publicly available registers. All prospective research, including essential details such as eligibility, sample size, interventions, and primary outcomes, must be registered, prior to enrolling patients. This is an important safeguard against investigators changing those details, without an openly available rationale. Clearly, sample sizes may change (almost always decrease) because of unforeseen circumstances, eligibility criteria may be adjusted, and even major outcomes may be changed. However, the initially planned primary outcome must always be reported, even if it is adjusted prior to analysis of the results. Any change in outcomes reported, or additional outcomes, after the results have been analyzed must also be clearly reported as being post hoc, and are always only hypothesis generating.

A recent study of a relatively non-invasive test of lung maturity (testing the L:S ratio on gastric aspirates (GAS), from a gastric tube that was inserted for clinical reasons) was performed, and some results have just been published by the clinical investigators. Heiring C, et al. Predicting Surfactant Need at Birth: Failed Validation of a Bedside Method Using Gastric Aspirates. Acta Paediatr. 2025. The method failed to adequately predict the need for surfactant… as far as we know.

This was a 4 centre study from Denmark among infants of <30 weeks gestation less than 45 minutes of age, who had not yet had surfactant. The primary outcome as noted on the registration page was the L:S ratio on the samples, the registration page notes : “The primary objective is to measure the L/S-ratio in fresh GAS using the AIMI 1.0/2.0 L/S POC Device and compare the L/S-ratio with the need for surfactant treatment aiming to validate the previously defined cut-off L/S-ratio for surfactant treatment and to determine if the cut-off L/S ratio needs adjustment before starting FAST 2 RCT”.

Of note, this is a really good acronym! It comes from the methodology used and the intervention expected Fourier trAnsform infra-red spectroscopy guided Surfactant Therapy. Or maybe, Fast Assessment of Surfactant deficiency to speed up Treatment.

Having performed the study, consented parents, and submitted the samples for testing, the patent holders who work with the company are refusing to allow publication of the pre-defined outcome. As the title of this post states, those partners in this study are clearly acting unethically, they are abusing the trust of the families who consented to the study.

Fortunately, we know about this because of the courage of the clinical investigators, and of the Editors of Acta Paediatrica, who have written and published the above article describing the study, documenting the dispute, and noting “disagreements over how the study findings should be reported and which findings to include. Specifically, the laboratory group proposed an unbalanced emphasis on lecithin (DPPC) alone as a predictor of surfactant treatment, based on post hoc analyses using an open dataset outside the framework of the agreed-upon protocol”. As I mentioned above, I don’t think that post-hoc analyses should be banned, but it is essential to focus on the results of the primary, pre-planned analyses. Anything that results from an inspection of the results, after they have been collected, is inherently unreliable, and must be submitted to further independent testing.

It appears that the company has “swivelled” and an abstract at the recent PAS-meeting appears to be reporting a study using the same device, but now discusses its use for predicting prolonged respiratory support (>6 h duration), not surfactant requirement. Firstly, I would caution the researchers at the Mayo clinic to ensure that they have a legally binding agreement to publish the results, especially the pre-specified primary outcome, otherwise they may find themselves in the same dilemma as the Danish researchers. Secondly, what is the purpose of this? How does that help? what would you do about it? The abstract doesn’t have enough detail to explain potential uses, but hopefully the investigators will make that clear in the future.

It is not unusual for a company which has invested in development of a new drug/technique/machine, and then finds it not to be very useful for the initial indication, to find something posthoc and swivel to that as an indication. It is an understandable reaction, one does not wish to lose the investment that has previously been made, and finding another indication might save the family jewels. One positive example of this is sildenafil, which was initially being developed as an angina treatment (where it seems to be effective, but may cause profound hypotension if the patient then takes their nitrates), but the company, post hoc, noted the frequent side effect of erections! And we all know what followed.

Research ethics approvals should include a legally binding agreement that the results will be submitted for publication, and that the publication must report the approved primary outcome, which must be identified as such. I don’t believe that is the case in many jurisdictions, but it should become the norm, to avoid situations such as this one.

Posted in Neonatal Research | Tagged , | Leave a comment

Caring for the most extremely immature infants

Posted on 29 May 2025 by Keith Barrington

There have been multiple publications concerning this issue recently, many from the tiny baby collaborative.

The first 2 publications are about the overall approach to providing intensive care at extremely low GA:
Bernardini LB, et al. It’s the little things. A framework and guidance for programs to care for infants 22-23 weeks’ gestational age. J Perinatol. 2025. This is a discussion of the many issues that should be addressed in centres trying to improve outcomes for these babies, including a recognition that they face specific challenges, require particular attention to detail, have unique physiologic limitations, and deserve an integrated caring approach with a committed team which includes obstetricians, nurses, all the allied health professionals, and parents. Indeed, despite the well-documented differences in the clinical approach of centres with good outcomes, one thing they all have in common is a belief that these babies can do well! You may have to convince your obstetricians that a delivery at 22 weeks is not a miscarriage, but is an extremely high-risk delivery that deserves the best possible care.

Many of the important considerations in this graphic are, appropriately, a bit vague. It is hard to disagree that maintaining “stable BP”, for example, is important, but exactly how to do that, and what to do when the BP is drifting downward at 4 hours of age, is beyond the scope of this graphic, and, unfortunately, has almost no evidence base to determine best practice.

Whatever you do, though, you should try to keep doing the same thing. Protocolized care (which means developing and following protocols) is essential in order to provide quality consistent care. (Al Gharaibeh FN, et al. The impact of standardization of care for neonates born at 22-23 weeks gestation. J Perinatol. 2025). This report from a health care regional program, treating 30,000 annual deliveries shows the results of the progressive implementation of a program to support the care of such babies. Guidelines covering many different aspects of care of these infants were implemented. In the early period, none of the 19 infants of 22 wk GA, and 39 of the 45 23-week infants received intensive care. This increased progressively to the post-implementation epoch to reach 29/34 and 48/51. For some reason these data are presented as Odds Ratios, which makes no sense to me when they are the result of an active decision. More importantly, survival improved, even when limited to the infants receiving active treatment, and complications of prematurity were stable or improved. Length of stay of infants receiving intensive care was shortened.

There are many publications documenting the efficacy of protocolizing complex care. The extremely immature infant is a prime example of a group who require such an approach.

Isayama T, et al. Survival and unique clinical practices of extremely preterm infants born at 22-23 weeks’ gestation in Japan: a national survey. Arch Dis Child Fetal Neonatal Ed. 2024;110(1):17-22 This paper is a good example of that premise, despite doing a lot of different things in different centres (some of which I would immediately label as “unnecessary”, “excessive” or even “dumb”), survival is very high. Japanese centres have strict protocols that all the staff follow, such that survival in this cohort at 22 weeks was 63% of those resuscitated (the majority) and at 23 weeks, 80% of those resuscitated (all except 2 of 757 infants). Among questionable practices in their protocols, 128 of the 145 level 3 centres in Japan perform echocardiography at least 3 times a day in the first 3 days, they measure a variety of different variables, as shown below, but what exactly they do in response to AV valve regurgitation, for example, is unclear.

116 of the centres also perform head ultrasound twice a day. What on earth you do about the head ultrasound result I also have no idea, especially as redirection of care is extremely uncommon in Japan.

Most of their ventilated babies are sedated with phenobarbitol or morphine, many also use fentanyl, even though it is useless as a sedative. 90% give probiotics, and some use donor milk if there is insufficient maternal milk. I was surprised to see that 50% give formula in this situation, but it seems to happen rarely in Japan, from other studies I have seen. Nearly all of them, 86%, give glycerine enemas in the 1st few days after birth, often on multiple occasions.

It is hard to argue with success, and the survival rates, of a largely unselected population are excellent. However, some of the longer term outcomes from Japan are concerning. This brand new publication, for example, (Haga M, et al. Prevalence and risk factors for neurodevelopmental impairment in very preterm infants without severe intraventricular hemorrhage or periventricular leukomalacia. Early Hum Dev. 2025;206:106286). Shows rather poor outcomes among the babies of 22 and 23 weeks, even though they are selecting, for this publication, the infants who do not have severe IVH or PVL. The results are not directly comparable to those from other countries as they use a Japanese evaluation tool, the Kyoto Scale of Psychologic Development, but the statistical spread of the results is similar to other tests, being normalised with a mean of 100 and an SD of 15. The following selection from their results includes the classification of the developmental test result first, with normal being >84, and delay being <70. You can see that the results are quite concerning at 22 and at 23 weeks; then, in the table, appear the actual mean scores and the incidence of CP.

This contrasts with outcomes from other places, such as these national data from Sweden, which include infants with brain injury on ultrasound (Soderstrom F, et al. Outcomes of a uniformly active approach to infants born at 22-24 weeks of gestation. Arch Dis Child Fetal Neonatal Ed. 2021). Of note, not all these infants had formal developmental testing, but Moderate-Severe is similar to the “delay” in the above study.

It may be just my prejudices, but I like to think that the much less interventionist approach in Sweden, with fewer ultrasounds, and more focus on integrating parents, helps to lead to better long term development.

To go back to the obstetric part, the following article confirms the marked lack of evidence to support any intervention for the mother threatening delivery at 22 or 23 weeks. (LeMoine FV, et al. Considerations for obstetric management of births 22-25 weeks’ gestation. J Perinatol. 2025). The weight of the limited observational evidence is strongly in favour of steroid administration, however, and probably also magnesium sulphate.

Agren J, et al. Tiny baby math: supralinear implications for management of infants born at less than 24 weeks gestation. J Perinatol. 2025. This article is an explanation of the major impact of the tiny size of these patients on everything we do to them. The relative volume of fluid flushes leads to very high potential sodium administration rates and possible seriously excessive heparin doses. We need to develop better small equipment, be prepared to use 2.0 mm ETTs for example, and to reduce the volume of blood taken for lab testing, we can completely eliminate CRP testing for example (that’s my take, not theirs) and we can run electrolytes exclusively on the same whole blood sample that we use for the blood gas with no increase in volume (and also measure ionized calcium, glucose, lactate, total bilirubin).

Which brings us neatly to fluid balance management, firstly an analysis of current use of humidification (Stoll CM, et al. Approaches to incubator humidification at <25 weeks’ gestation and potential impacts on infants. J Perinatol. 2025) which shows that centres start at varying relative humidity, mostly over 75%, because there is no good evidence to decide what to start at, but that weaning can probably be quite fast. Using high humidity will help to avoid major trans-epidermal fluid loss, and the accompanying heat loss, from the latent heat of vapourisation (high-school physics!) and the consequent hypernatraemia.

That is confirmed by this study from Upssala (Naseh N, et al. Fluid Balance in Infants born at 22-23 Weeks’ Gestation: Trajectories and Associations with Outcomes. J Pediatr. 2025:114661), where they commence incubator humidity at 85%, and start IV fluids at 110 mL/kg/d at 23 weeks, or 120 mL/kg/d at 22 weeks. They adjust subsequently to aim for the following a) Maximum weight loss of 10–15% at a postnatal age of 3-5 days; b) Weight deficit of ~10% at 7 days, and regain of birth weight by 10–14 days; c) Plasma sodium <150 mmol/L; d) Initial sodium intake <4 mmol/kg/d. Which are similar to our goals, and I would think, many other centres.

I really don’t like the way this graphic is constructed, for one thing, I think the dotted lines, which are labelled as “Incidence (%)” are actually prevalence, i.e. the proportion of infants on that particular day with that diagnosis (if it was incidence it couldn’t go down again), and putting a continuous variable like weight loss on the the same graph as a discrete variable, like the proportion of a relatively small number of patients, which is then shown as a line, is really questionable. Acute Kidney Injury, here is defined by oliguria. Of the 69 included infants, 7 received insulin, the precise indications for which I couldn’t find, but there were more than 7 who had a blood glucose >20 mmol/L, so I presume it must be a persistent blood glucose >20.


As you can see from the error bars, weight loss of >15% was not rare, and was associated with increased mortality. They don’t mention IVH, but in our local data (which I haven’t yet published) the association of hypernatraemia (>145) with concurrent hyperglycaemia (>12 mmol/L) often led to severe hyperosmolarity, with peak osmolarity often >320, which was strongly associated with severe IVH.

Looking after these babies is a real challenge, many centres are now reporting survival at 22 weeks which is greater than 1/3, and in Japan is approaching 2/3, with dramatically better survival at 23 weeks, Compared to the rest of care for seriously or critically ill patients of other ages, that is very far from being futile! The quality of life of the large majority of survivors is excellent, even though many have challenges, in particular with speech development and executive function. Those challenges should be used as a lever to improve educational and other support services for these patients, not an excuse to deny them intensive care.

It seems that more and more centres are offering active intensive care to infants at these profoundly low gestational ages. I think it is often appropriate to give these babies a chance to “prove themselves”, but we must take into account other risk factors, in particular growth restriction and outborn status, which is often accompanied by lack of antenatal steroids, when we discuss the best approach with parents.

We owe it to these families to do everything we possibly can to improve their chances, with a dedicated team, who believe that these babies are worth the effort. A team which has examined the literature and their own practices in detail, have constructed clear protocolized care plans, and are prepared to follow them. Only then can these most immature babies get the care they deserve.

Posted in Advocating for impaired children, Neonatal Research | Tagged , , , , | 3 Comments

Hypotension and Shock. Optimising treatments

Posted on 27 May 2025 by Keith Barrington

A new single centre RCT of permissive hypotension (PH) compared to “standard treatment” (ST) of very preterm infants 24 to <30 weeks GA, with a mean BP lower than their GA has just appeared (Alderliesten T, et al. Treatment of Hypotension of Prematurity: a randomised trial. Arch Dis Child Fetal Neonatal Ed. 2025). In the intervention, PH, group, infants only received treatment if they developed signs of poor perfusion. In the ST group they immediately received a fluid bolus and a dopamine infusion, followed by adding dobutamine, then epinephrine, hydrocortisone was given if the baby needed more than just dopamine. They don’t explicitly say what the goals of the cardiovascular support were, and what triggered increases in dose or the addition of other agents. I assume that the goal was to have a mean BP above the GA, but perhaps the goal was the GA+3 or something, it should have been stated. About 1/3 of the 40 ST babies only had a fluid bolus, and no further catecholamine support, about 1/4 of the 46 PH babies eventually had a fluid bolus. From table 2 it looks like there were 19 babies who had a fluid bolus and went on to receive catecholamines in the ST group, and 6 PH babies, at least, if the numbers in the table were the maximum doses received.

The table is poorly explained, the floating numbers in the right-most column are presumably p-values, and I assume that the first of them (0.004) refers to the whole block of data about what seems to be maximum intervention received, and not to the 0 vs 0! It seems that only 32 of the 40 ST babies actually received any treatment for their hypotension.

The babies were enrolled into the study between 2011 and 2018, and the study was eventually terminated, in part because of falling eligibility. This period probably covers the introduction of routine delayed cord clamping in preterm deliveries, which has led to a major reduction in the diagnosis and treatment of hypotension.

The interventions in the PH group presumably are for 15 babies who developed signs of poor perfusion, or a mean BP of GA-5 mmHg (according the discussion section they were mostly for low BP); but 12 of them had good perfusion after a fluid bolus, and they had no further intervention. Why there are, it seems, 8 ST babies who didn’t receive any intervention isn’t clear. What you can also see in this publication, is that the mean BP in the 2 groups was just about identical, despite the interventions in the ST group.

The primary outcome of the trial was developmental outcome using Bayley version 3 cognitive and motor scores at 24 months corrected age. Secondary outcomes included mortality and the usual NICU complications.

The major finding of the trial is that the 2 groups had almost identical long term outcomes. The mean scores, and the proportion <-1SD below the mean, and traditional combined outcomes, were all very similar between the groups, as were the acute complications, NEC and IVH. The mortality was slightly higher in the permissive hypotension group 6/46, vs 3/40, but they note that there were 2 PH babies who died after the intervention period because of LOS with G-negative organisms and shock.

I do not understand the presentation of the cognitive and motor scores; they are identical between the 2 study arms, but they are presented as being approximately 91 in the 2 groups. But, further down in the main results table, they are presented as “using age at BSID-III assessment corrected (CA) for prematurity” and the means are closer to 101. Surely the primary outcome was already the BSID scores corrected for prematurity?

The interventions, therefore were different to our HIP trial (Dempsey EM, et al. Hypotension in Preterm Infants (HIP) randomised trial. Arch Dis Child Fetal Neonatal Ed. 2021;106(4):398-403). In our trial the infants in both groups were also free of signs of poor perfusion, and they all received a fluid bolus, then the intervention group were started on 5 microg/kg/min of dopamine, compared to placebo in the controls. This meant that the BPs were different between groups:

We have recently published our long term outcomes (Marlow N, et al. Outcomes of extremely preterm infants who participated in a randomised trial of dopamine for treatment of hypotension (the HIP trial) at 2 years corrected age. Arch Dis Child Fetal Neonatal Ed. 2025) which showed no major differences, but with the proviso that we were very underpowered, and terminated the trial well before the planned sample size.

There was, if anything, a trend towards a better outcome in the intervention group.

In the new trial, one can really quibble about the interventions used. Although dopamine is most commonly used for this indication around the world, which is why we used it in the HIP trial, it acts as a pure vasoconstrictor in the preterm infant, improved perfusion of any organ has not been shown with dopamine use in the newborn. In particular dopamine is a cerebral vasoconstrictor, and has been shown in several models to decrease brain blood flow, or brain oxygenation.

Second line therapy with dobutamine is also debatable, dobutamine is a vasodilator, and does not reliably increase BP, so using it to treat hypotension is questionable (although it may be effective in improving perfusion).

There are a few concerns with this article, the delay between completion of follow up (2020) and publication is very long, the specific items that I have mentioned above, but most importantly they didn’t reference the original article which invented the term “permissive hypotension”! Dempsey EM, et al. Permissive hypotension in the extremely low birthweight infant with signs of good perfusion. Arch Dis Child Fetal Neonatal Ed. 2009;94(4):F241-4.

The sad fact is that we still don’t know what to do about the baby with reasonable clinical perfusion who has a numerically low BP. The totality of the current data suggests that it is entirely acceptable to just wait and see, especially in the current era of delayed cord clamping, where hypovolaemia is very unlikely.

Posted in Neonatal Research | Tagged , , , | Leave a comment

Supporting Breast Milk production in the NICU

Posted on 26 May 2025 by Keith Barrington

A recent confluence of 3 reviews, one on breastfeeding specifically in the NICU (Hilditch C, et al. Effect of Neonatal Unit Interventions Designed to Increase Breastfeeding in Preterm Infants: An Overview of Systematic Reviews. Neonatology. 2024;121(4):411-20) another on the evidence supporting the efficacy of lactation consultants (D’Hollander CJ, et al. Breastfeeding Support Provided by Lactation Consultants: A Systematic Review and Meta-Analysis. JAMA Pediatr. 2025;179(5):508-20), and the US Preventive Services Task Force report (Preventive Services Task Force. Primary Care Behavioral Counseling Interventions to Support Breastfeeding. JAMA. 2025).

There is some overlap between the reviews, and also some important differences. The USPSTF report is primarily concerned with primary care, but, of course, most of the mothers of NICU patients start out there. The final recommendation is a bit vague, but supports interventions to increase breastfeeding, either with personal counselling, or referral to a breast feeding specialist. The recommendation is based on a total of 90 RCTs, including nearly 50,000 women, which studied a wide range of interventions. Most of the studies were in high income countries, especially the USA, where about 60% of babies are breastfed at 6 months, and 42% are exclusively breastfed.

It is hard to believe that any mother in 2025 is unaware of the benefits of breast feeding, so simply informing a pregnant woman or a new mother of the advantages of breast milk is clearly not enough. As an accompanying editorial notes “clinicians and health systems must do more than just promote the benefits of breastfeeding and provide basic education. Behavioral breastfeeding education and counseling should be considered a starting point within a broader system that requires retooling, rethinking, and innovating”. That is particularly so within the USA, which is one of only 7 countries in the world with no nationally mandated paid maternity leave, the others being The Marshall Islands, Micronesia, Nauru, Palau, Papua New Guinea, and Tonga!

The USPSTF definition of breast feeding includes provision of expressed breast milk, which is so common in the USA; many new mothers returning to work very quickly after delivery, even before breast feeding has been well established. It is bizarre that so many articles, about breast feeding support in the USA, focus heavily on helping new mothers to find privacy, support, time, and equipment for breast milk expression, so that someone else can give their breast milk to the new baby. Please don’t get me wrong, I have nothing against a mother who makes the free choice to do that; but most new mothers in the USA have little choice, they have to rapidly return to work or risk losing income or even their job.

The review of the efficacy of Lactation Consultants also uses the definition of breastfeeding as provided in the original publication, 22 of the 40 RCTs they included were performed in the USA, so will have included feeding expressed breast milk. The review demonstrates that a referral to a lactation consultant is effective in prolonging breast feeding by an average of 3.6 weeks. When the intervention commenced sooner after birth, the metaregression showed a greater impact.

Within the NICU, interventions shown to improve the proportion of mothers breast-feeding at discharge are: skin-to-skin care (but there is no clear additional advantage to starting skin-to-skin care within 24 hours); and avoiding bottles during the establishment of breast feeding, very low quality evidence supported cup-feeding as the alternative method of choice. For breast feeding rates post-discharge, the evidence base was even poorer, but supports the same interventions. What is exactly meant by “establishment of breast-feeding” is not consistent among trials, so the point at which bottles could be introduced without having a negative impact is unclear. In our NICU we try to delay the introduction of bottle feeds until there are at least 3 full nutritive breast feeds per day, and we currently use very little cup-feeding. Obviously, everything is done in partnership with the parents, who sometimes request a different approach, with some wanting bottles sooner, and others preferring their avoidance entirely.

This review, and the RCTs underlying it, do not address what is a more important outcome, at least in terms of acute NICU complications, and that is “how can we improve the amount of maternal breast milk that very preterm infants receive”? One possible adverse impact of breast milk banks, for example, is a reduction in the amount of maternal milk provided, a systematic review from a few years ago suggested that that happens sometimes, but not consistently, in some units the provision of mother’s milk increased. What is consistent, of course, is a reduction in artificial formula use. One of the best things we can do to reduce NEC, and probably LOS and mortality, is to increase the proportion of maternal breast milk received by our very preterm babies. More research in how to do that could have major benefits.

Whether prenatal intervention by a lactation consultant, for example, increases the proportion of MoM received has not been well studied, although I think it is likely to be very helpful. Observational data shows an improvement in breast milk production if milk expression is commenced very early after delivery, depending on the study this could be within 1, 2 or 3 hours. But what if we expended a small amount of resources to specifically ensure that someone with breast feeding support expertise met all the mothers of preterm infants within that 1 to 3 hour window, to ensure that they knew what to do, how to access the equipment, and how to ensure that the colostrum produced was sent immediately to the NICU? I think a randomized trial could be very useful to show whether or not that was beneficial. Many other aspects of breast-feeding support could also be better studied, so that we can develop best practices for an intervention which may not be “sexy”, but has a big impact on clinical outcomes of our patients.

Posted in Neonatal Research | Tagged , , , | Leave a comment

NEC awareness day, 17 May 2025. What is new in NEC prevention?

Posted on 17 May 2025 by Keith Barrington

I thought, as today is NEC awareness day, I would give an overview of recent research about NEC prevention strategies.

Probiotics.

I am sure my readers are aware of the major problems in the USA which have followed the occurrence of an episode of sepsis caused by a probiotic organism; with several preparations being no longer available. This has happened even though the data generally shows a reduction in late-onset sepsis overall with probiotic supplementation, or in some analyses, no difference in overall LOS frequency. I’m not aware of any study, or any systematic review, which has shown an increase in sepsis with probiotic prophylaxis.

We have just published our experience with sepsis caused by organisms in the probiotic mixture that we were using (Abda A, et al. Bifidobacterium And Lactobacillus Bacteremia among Infants Receiving Probiotics in the Neonatal Intensive Care Unit. J Pediatr. 2025:114521). Among over 2000 infants who received probiotic prophylaxis in our NICU there were 12 cases of bacteraemia with one of the bugs in the mixture. Over that period we had 2 other babies who did not fit criteria for the study, who had bacteraemia with either a lactobacillus or a bifidobacterium that was either a different strain to our mixture or in a baby who had never received the mix.

The babies who had bacteraemia frequently had major GI problems, especially perforation, which creates, obviously, an entry portal for the probiotic organisms to invade the circulation, the affected infants were all less than 28 weeks gestation. The degree of illness was mild among those without other serious complications, (clearly an extreme preterm infant with GI perforation is often very sick). No mortality was attributed to the sepsis.

There is also a new analysis from the CNN, (Alshaikh BN, et al. Effectiveness and Risks of Probiotics in Preterm Infants. Pediatrics. 2025;155(3):e2024069102), confirming the benefit of probiotics in current usage in Canada. This very large observational study showed lower mortality with probiotic usage, and a limited reduction in NEC. For the purposes of this discussion, they showed a small reduction in late-onset sepsis. Despite an incidence of LOS caused by probiotic organisms of about 0.4% (not very different to our local rate of 0.6%) the overall adjusted Odds Ratio of LOS was 0.9.

A brand new SR of probiotic sepsis (Feldman K, et al. Incidence of probiotic sepsis and morbidity risk in premature infants: a meta-analysis. Pediatr Res. 2025) calculates the frequency as being much lower than that, at 8/20,000 (but depends on how many cases were reported in the published RCTs). It calculates the risk benefit as overwhelmingly in favour of probiotics, which is still the case even if the true incidence of probiotic sepsis is 10 times higher than their estimate.

Prebiotics

There is a huge amount of ongoing research into prebiotics, most particularly the Human Milk Oligosaccharides (I just did a PubMed search on “HMOs and newborn” and there are already 120 articles in 2025). One of the challenges with probiotic usage in the preterm is the difficulty in achieving an intestinal microbiome which is similar to healthy, vaginally delivered, exclusively breast-fed term infants. Despite colonization with probiotic organisms, they only represent a small proportion of the gut bacteria. One potential way of improving this (which has been shown to be effective in other situations, such as after antibiotics) is to co-supply probiotics and HMOs. This is, of course, what breast feeding does! (Lubiech K, et al. Breast Milk as a Source of Prebiotic Human Milk Oligosaccharides and Bacteria from the Lactobacillaceae Family. Folia Biol (Praha). 2025;71(1):44-53). Recent articles have shown an impact on gut colonization after antibiotics of prebiotic supplementation, and, in this mouse model of NEC, HMOs seemed to decrease the loss of enteric neural cells. Whether this was a direct effect, or mediated by changes in the microbiome isn’t clear to me, as they don’t seem to have done any microbiome analysis. (Sodhi CP, et al. Necrotizing enterocolitis: specific human milk oligosaccharides prevent enteric glia loss and hypomotility. Pediatr Res. 2025).

There is clearly much work yet to be done, but it seems likely to me that supplementing enteral feeding with human milk, preferably unpasteurized maternal milk, with a probiotic mixture and HMOs to support growth of the organisms, especially, probably, with the use of DSLNT, will allow much more appropriate intestinal colonization, and our best chance of reducing NEC.

Breast Milk

Another analysis from the CNN reinforces much of what we know about NEC, and introduces something that we haven’t discussed much, which is the period of time that an infant is npo after birth. (Bando N, et al. Association of Enteral Feed Type with Neurodevelopmental and Neonatal Outcomes among Infants Born Preterm. J Pediatr. 2025:114536). This is an analysis of infants <29 weeks GA who received either maternal or donated human milk, or a mixture, but no formula, within the first 28 days of life, and did not have more than 7 days npo during this period. There is a lot of information about follow-up in this article, which supports an association between maternal milk (Mother’s own Milk MoM) and improved long term outcomes, such as cognitive scores, compared to donor milk, or a mixture, or being npo.

For the purposes of this blog post, there is also a clear association between being npo and NEC, with each extra day of being npo increasing the Odds Ratio for NEC, or conversely, as shown here, an OR of about 0.77 for each day of receiving enteral human milk feeds rather than being npo. For this outcome in this analysis, donor milk was very similar to MoM.

There is often a period of being npo when an at-risk baby is admitted to the NICU, as we get lines inserted, stabilise the respiratory status, and so on. There is also, often, little breast milk production, sometimes for very good reasons of maternal illness, but sometimes because the post-partum nurses are not as focused on breast milk production as we are. It is very important to make this period as short as possible, and, as long as the baby is not in shock, we should start to give breast milk within the first couple of hours of birth. Developing systems to ensure that all MoM produced finds its way immediately to the NICU, so that we can administer to the infants, is essential. This new analysis suggests that donor milk might be just as good as mother’s milk in terms of reducing the risks of being npo; but overall, I think it is clear that MoM is preferable.

Oro-pharyngeal Colostrum

I have already blogged about my scepticism of the benefits of this procedure for preventing NEC. A new meta-analysis, which is supposed to be highly supportive, includes many trials of questionable relevance. (Anne RP, et al. Effect of oropharyngeal colostrum therapy on neonatal sepsis in preterm neonates: A systematic review and meta-analysis. J Pediatr Gastroenterol Nutr. 2024;78(3):471-87). One of the trials, which contributes 48% of the weight of the meta-analysis of the impact on NEC (Mannan 2023), had a 65% incidence of stage 2 NEC, among infants who were almost all >28 weeks GA! And includes yet another moderately large Chinese trial which was retrospectively registered, and published in contravention of the guidelines of the journal where it appeared (OuYang 2021).

The effects of the intervention on LOS were also heavily influenced by a single trial, weight 38%, which only included infants >28 weeks, and had an incidence of LOS of 75% (reduced to about 60% by the intervention. Lenonardo 2022). The inclusion of data from centres with outcomes so wildly different to our current practice in Canada, make their relevance highly questionable to my NICU.

One advantage of trying to administer colostrum very rapidly to our very preterm babies, is that it requires that the logistics of ensuring that the precious liquid arrives at the incubator-side of the baby very quickly are all in place. This should help us to reduce the time that a baby is npo to a minimum. Painting the inside of the mouth with colostrum is probably harmless, but please put the rest of the breast milk into the GI tract!

Reducing antibiotic use

An infant born by cesarean section without labour and with intact membranes does not require antibiotics. The chances of sepsis are zero. Many other preterm babies can either avoid, or have very short duration (36 hours), antibiotic therapy. The longer a baby is treated with antibiotics the greater the risk of later development of NEC, as confirmed, again, by this secondary analysis of the PENUT trial (Strobel KM, et al. Effect of early antibiotic exposure on necrotizing enterocolitis and growth in extremely preterm infants. Pediatr Res. 2025).

BreastMilk Fortification

And finally, there is still no evidence that the timing, or the type, of fortifier, added to breast milk has an impact on NEC occurrence. Once you have stopped giving artificial formula to at-risk babies, there is no evidence that one fortifier compared to another, whatever the source of the original components, changes clinical outcomes.

Other interventions

It remains unclear how blood transfusions and severe anaemia are related to NEC. Recent RCTs which have either examined transfusion thresholds, or have reduced transfusions by giving erythropoietin or analogues, have not shown a difference in NEC. It is possible that more severe anaemia than allowed by the protocols of those trials could be associated with NEC.

Gastric acid suppression can be entirely avoided in at risk patients, there is no clear indication for their use in these babies.

How we treat the PDA has little impact on NEC, although the Cochrane review shows that ibuprofen leads to less NEC than indomethacin.

On this NEC awareness day, we must remain committed to the need for more research, and for more high quality studies, to eliminate this scourge from our NICUs.

Posted in Neonatal Research | Tagged , , , , , , , , , | Leave a comment

Should we be giving erythropoietin, or analogues, to preterm infants?

Posted on 14 May 2025 by Keith Barrington

In some pre-clinical models, erythropoietin acts as a neuroprotective agent, which led to the performance of clinical trials to determine whether there was a positive impact on outcomes of the preterm. The latest study has just been published, and it is another high quality trial from the NICHD NRN, and Robin Ohls (Ohls RK, et al. Darbepoetin, Red Cell Mass, and Neuroprotection in Preterm Infants: A Randomized Clinical Trial. JAMA Pediatr. 2025). The article starts with a questionable assertion, “Erythropoiesis-stimulating agents (ESAs) such as erythropoietin…have shown evidence for neuroprotection in preterm infants”. I think if that was true, they would not have done this trial! In fact the references they give for that statement show a very variable impact on outcomes, the figure below is from a SR/MA which included only erythropoietin studies, eliminating a darbepoietin (a long acting EPO analogue) group in the second study by Robin Ohls et al (2014).

This is not a criticism of this trial, or of the justification for performing it, which I think was very well founded, just the way that sentence is worded.

It is very important to note that the possible benefit of EPO on long term outcomes is entirely dependent on Song 2016, which was retrospectively registered after completion, and published in contravention of the guidelines of the journal where it was published. If you delete the Song data there is very little indication of a benefit of EPO on the long term.

The highest quality trial among those analyzed in the SR/MA was Juul, which showed nothing, in terms of impacts on long term neurological or developmental outcomes. The Natalucci, Swiss EPO trial, was also high quality, and about 2/3 the size, and was also without long term benefit.

The new trial enrolled 650 infants between 23 and <29 weeks GA, <36 hours of age, and continuing until 35 weeks PMA. Darbepoietin 10 mcg/kg or placebo was given weekly, IV while the access was in place followed by subcutaneous Darbe, or a sham procedure as placebo.

The primary outcome, I am very happy to say, was NOT “death or NDI”!!!! The primary outcome was a continuous variable, the Bayley version 3 cognitive composite, with dead infants being assigned as score of 54, which is the lowest possible score. I’m not sure this is the best approach, as the difference between a score and 55 and 56 is the the same as between 55 and being dead. As there was no imbalance in the deaths (50 in each group during hospitalisation, and 2 Darbe vs 3 controls after discharge) it probably changes little how exactly the deaths were accounted for, in terms of the difference between groups.

On the other hand, the scores seem lower than they are if you calculate them solely among the survivors, which results are given in the supplementary appendix.

And the higher mortality among the more immature group has a greater impact on the apparent cognitive scores in that group

Even though hidden away in the supplementary appendix, that is the result that I think is most interesting; among survivors what was the impact of darbepoietin on cognitive development at 2 years corrected age?

The supplementary appendix also shows that the other components of the Bayely scores were similar between groups, and that there was no difference in neurological impairment (cerebral palsy or visual or hearing impairment).

As for other impacts of darbepoietin prophylaxis, the intervention was very effective. It stimulated bone marrow red cell production, increasing haemoglobin, reducing transfusions (with a transfusion guideline which was the same, of course, for the 2 groups), and reducing donor exposure.

This is part of the figure showing that darbepoietin, yellow circles, increased haematocrit compared to the blue circle controls, despite fewer transfusions.

Darbepoietin doubled the proportion of babies who never had a transfusion, decreased the total number and volume of blood transfused, and decreased the number of donor exposures.

It is also interesting that this did not affect the frequency of surgical NEC, and there was less severe BPD, and probably less severe retinopathy.

The lack of impact on surgical NEC is interesting and consistent with all the other studies, and is another evidence point against the existence of so-called TRaGI, Transfusion-associated Gut Injury. To explain, there is an idea that transfusions trigger NEC in preterm infants, and that how we feed infants around blood transfusions might have an impact on the complication. But if you look at all the data, there is no evidence from prospective studies of transfusion thresholds, or of EPO and similar agents, that fewer transfusions leads to less NEC. I think that is the best evidence we have that the association between transfusion and NEC that has sometimes been reported is just coincidental.

The reduction in severe BPD was not seen in the PENUT trial, nor in the large Swiss EPO trial, so might well just be a random finding, but, at the very least, there is no evidence of increased lung injury with these agents. Similarly with retinopathy, there is generally a tendency to have somewhat fewer cases.

I’ve never seen an analysis of the cost benefit of these agents, but the dose we give to babies is tiny (a quick search looks like it costs about 3 dollars per microgram in Canada), and the costs of transfusion are substantial, even though blood in most countries is donated freely by heroes, the total cost of a transfusion in Canada was recently estimated to be about $500. I think it is likely that darbepoietin is cost-effective, and, if I was a parent in the NICU currently, and given the choice of darbepoietin compared to an increased risk of transfusion, I think I would choose darbepoietin. The only downside is the pain of a weekly subcutaneous injection, which is the main advantage of darbepoietin, compared to erythropoietin, which requires painful administration 3 times a week.

I think that based on all these recent data, that erythropoietin and analogues are safe, in the short and the long term, effectively increase red blood cell production, and reduce blood transfusion in a clinically significant way. Their use should be standard of care for very preterm infants, or offering their use to parents should be standard practice. Just don’t expect them to improve neurological or developmental outcomes

Posted in Neonatal Research | Tagged , , , , , , | Leave a comment

Neonatal Research Shorts : May 2025

Posted on 13 May 2025 by Keith Barrington

Dutta S, et al. Seven-day versus 14-day antibiotic course for culture-proven neonatal sepsis: a multicentre randomised non-inferiority trial in a low and middle-income country. Arch Dis Child Fetal Neonatal Ed. 2025.

This is a multicentre trial from India of neonates (0-28 d of age), term and preterm (down to 1kg b.wt) who had culture positive sepsis and were stable by 5 days of treatment. There is a detailed algorithm in the supplemental materials to define “clinical remission of sepsis”, which looks quite fair. At 7 days of treatment of sepsis, with antibiotics to which the organism was shown to be sensitive, the babies were randomized to stopping the antibiotics, or continuing for 14ays. They excluded babies with S Aureus, or fungi, or meninigitis, or other infections considered to definitely need longer antibiotic courses, such as osteomyelitis. There were about 1000 babies with culture positive sepsis, and following the deaths, transfers and non-eligibility criteria there were 261 randomized.

The primary outcome of the trial was definite or probable relapse, with definite being defined by recurrent infection with the same organism, and probable being “culture-negative sepsis”, within 21 days of the end of the antibiotics. The short duration antibiotic group could be discharged before the end of the antibiotic course of the long duration group, and most of them were, the median hospital stay after randomization was 6 days in the short treatment group. That introduces a possible bias, as surveillance, and risk factors, are obviously different after hospital discharge. Also, many of the results, including the primary outcome is calculated as the occurrence of the outcome within 21 days after stopping the antibiotics. In other words, a 7 day longer period in the long treatment group than the short duration group.

It looks like someone during the peer-review process had the same concern, and the results are all also presented as the occurrence of the outcome during various time periods after randomization. You can see those results below. The first column being the 7-day duration group, the2nd being the 14-day group.

There are no important differences, with all outcomes being in favour of the short treatment, and the trial was stopped early, after about 2/3 completion, because the DSMC stated they had demonstrated non-inferiority. The duration of hospitalisation after randomization was 4 days shorter in the short treatment group, despite the inclusion of preterm infants whose discharge may well be unrelated to the sepsis treatment.

Are these data relevant to High Income Countries? The bacteriology is somewhat different to HICs, this study had no cases of GBS, and a variety of mostly Gram-negative organisms, including Acinetobacter baumanii. Many of the organisms are multi-resistant, including the A. baumanii which is always Beta-lactam resistant, and is usually resistant to aminoglycosides and to meropenem. (Russell NJ, et al. Patterns of antibiotic use, pathogens, and prediction of mortality in hospitalized neonates and young infants with sepsis: A global neonatal sepsis observational cohort study (NeoOBS). PLoS Med. 2023;20(6):e1004179).

Take Home Message : It is reasonable to consider shorter courses of IV antibiotics in newborns with culture positive sepsis who have received 7 days of treatment with antibiotics to which the organism is sensitive, if they stabilised, and were clinically improved, within the first 5 days of treatment.

Hough JL, et al. Intermittent sigh breaths during high-frequency oscillatory ventilation in preterm infants: a randomised crossover study. Arch Dis Child Fetal Neonatal Ed. 2025;110(3):297-302.

I remember when we first started using HFO in preterm infants, we thought it might be important to have intermittent sighs, as Alison Froese had shown in an animal model that they seemed to be important for aiding lung expansion. She had shown that lungs which were atelectatic, such as we see in HMD, couldn’t really be opened up by HFO without sigh breaths. They rather fell out of use as we gained experience with HFO, but there are a number of studies looking at various methods for improving lung recruitment, including substantial increases in mean airway pressure, followed by rapid weaning. Like many others, given the lack of any evidence that first intention HFO improves pulmonary outcomes compared to first intention conventional IMV, I keep HFO to use as a rescue mode, and if I am changing to HFO because of problems with oxygenation (rather than difficulty with ventilation) I increase the mean airway pressure by 2 cmH2O initially, which often leads to a decrease in FiO2, and therefore, likely, an improvement in lung recruitment.

In this randomized cross-over trial, a very diverse group of preterm infants on HFO with an oxygen requirement (25 to 75% O2), who were already on HFO either stayed as they were, or had sigh breaths added, and then switched to the other condition. The sigh breaths were delivered with a pressure of 30 cm H2O, and an inspiratory time of 1 second, 3 times per minute. One part of the protocol I’m not sure I understand is that “MAP-set was reduced in proportion to the difference between MAP-set and PIP = 30 cmH2O aiming to keep average MAP unchanged”. I think I know what they did, its just the “…in proportion to the difference between…” part that is confusing. I think that, for an infant starting on an MAP of 15 cmH2O, during the sigh breath period, they would reduce the HFO mean pressure to account for the 3 seconds/min of a pressure of 30 cmH2O, in other words, they reduced the mean pressure to 14.2 during the sigh period. Or an infant on a MAP of 12 would be reduced to 11. They could just have said, “the set MAP was reduced slightly, to keep average MAP unchanged”. Each period lasted 2 hours. During the 2 periods, they estimated changes in lung volume using trans-thoracic impedance, and distribution of ventilation using the same methodology.

Although each period lasted 2 hours, they only present data at 0, 30 and 60 minutes, I am not sure why. Lung volumes appear to have increased during both phases of the study, but with a greater increase when sighs were added.

Take Home Message : Intermittent sigh breaths improve some regional lung inflation, and have a small effect on oxygenation. Further study should determine whether this improves pulmonary outcomes.

Muts J, et al. Macronutrient concentrations in human milk beyond the first half year of lactation: a cohort study. Arch Dis Child Fetal Neonatal Ed. 2025;110(3):248-52.

This was an analysis of milk donated to the Dutch national milk bank. Mothers who donated more than 6 months after delivery had the macronutrient composition of their donation. Protein concentrations decreased slightly up to 8 months of donation, and then re-increased. Carbohydrates remained stable, and fat concentration increased after 8 months.

Take Home Message : Up to 2 years after delivery, donated breast milk is satisfactory as a source for preterm infants.

Eisenberg MA, et al. Epinephrine vs Norepinephrine as Initial Treatment in Children With Septic Shock. JAMA Netw Open. 2025;8(4):e254720.

A retrospective single centre cohort study, from Boston Children’s, with over 200 children with septic shock, compared renal outcomes between those treated initially with epinephrine to those who received norepinephrine first. The primary, renal, outcomes were not different between groups, but survival was greater with norepinephrine, (0% mortality, compared to 4% mortality with epinephrine).

Take Home Message : we have very little reliable data on which to base choice of agents for cardiovascular support in septic shock. This study suggests that norepinephrine is a reasonable first choice, for older children.

McLean MA, et al. Neonatal Sucrose and Internalizing Behaviors at 18 Months in Children Born Very Preterm. JAMA Netw Open. 2025;8(4):e254477.

In 3 Canadian NICUs, daily pain exposures and sucrose administration were recorded for infants born at 24 to <33 weeks. Opiates and other analgesics were also recorded. 200 babies had Child Behaviour Checklists completed.

I was shocked to see that one of the 3 centres does not use Sucrose! I can’t imagine why. Sucrose clearly reduces pain in very preterm infants, and the idea of performing a mean of 170 painful procedures on the poor babies in centre 1 without the use of sucrose, I find appalling. For many painful procedures, such as heel puncture, opiates are ineffective, and pain control is optimal with a combination of sucrose, soother, and skin to skin care.

There was some older data from a trial by Celeste Johnston that suggested that the infants with the highest sucrose exposure had poorer long term outcomes, but they also had more painful procedures. No other data are available to show adverse long term impacts of sucrose.

The study showed that, “after accounting for site differences in clinical factors, greater neonatal pain (number of painful procedures) was associated with greater internalizing at 18 months CA. Cumulative sucrose exposure in early life was not associated with child internalizing behaviours”. They also note “point estimates indicate no clinically meaningful difference in internalizing behaviours for children exposed to sucrose (B = 2.38)”, and “Cumulative sucrose was not associated with internalizing scores (B = 0.41; 95% CI, −0.52 to 1.41) accounting for neonatal pain”. The authors then try and stretch their data, suggesting that, because the confidence intervals around the effect of sucrose on internalizing behaviour are asymmetric, the “point estimate” of the association between sucrose and the internalizing T-score on the CBCL is consistent with a clinically relevant impact.

Take Home Message : Painful procedures among preterm infants in the neonatal period are associated with increased internalizing behaviour at 2 years of age. Sucrose does not change that association.

Posted in Neonatal Research | Tagged , , | Leave a comment

Therapeutic Hypothermia at 35 weeks

Posted on 11 April 2025 by Keith Barrington

In response to a couple of thoughtful comments to my previous post, I thought I would try to address the specific issue of the infant of 35 weeks. (Faix RG, et al. Whole-Body Hypothermia for Neonatal Encephalopathy in Preterm Infants 33 to 35 Weeks’ Gestation: A Randomized Clinical Trial. JAMA Pediatr. 2025;179(4):396-406).

There is clearly some room for discussion, not only are obstetric estimates of gestational age uncertain, but what is the real difference between an infant at 35 weeks and 6 days, compared to 36 weeks exactly? Should we really determine therapy based on whether the baby was born before or after midnight?

The discussion section of this new article reports on personal communication with the investigators behind the 2 previous trials that included some babies at 35 weeks gestation. Apparently, in total there were 7 babies, 2 controls (1 died and 1 survived without handicap), and 5 cooled (2 deaths, 2 survivors with disability, and 1 without). From the supplemental data supplied with the new article, there were 48 infants of 35 wk GA in this trial, 20 controls and 28 infants treated with hypothermia.

Adding all the results together gives the following :

Control  Hypothermia  
DeathDisabilityNo disabilityDeathDisabilityNo disability
41175223

I performed an ad hoc Bayesian analysis myself, using an on-line calculator, which gave probabilities of close to 50% for each of the 2 options being preferable, for the combined outcome of “death or disability”, but you don’t really need to do that, there clearly is next to no difference in the outcomes.

This is, of course, an extremely limited analysis. I do not know if the definitions of disability (moderate or severe) were the same in those 2 prior trials as in the new trial, the 2 prior trials being Eicher et al and ICE. I don’t know how definite the obstetricians were in their calculation of GA, I don’t know how many babies in the various trials were lost to follow up, or if there were babies who died between discharge and follow up in the previous trials. All that said, the very limited data available do not show a major benefit of cooling at 35 weeks.

We could ask, why would therapeutic hypothermia suddenly become effective at 36 weeks 0 days? That is inherently unlikely, and I think we really need to have an Individual Patient Data meta-analysis, analysing the outcomes of the hypothermia trials by week of gestational age, the most likely outcome of such an analysis, I posit, would be a progressively greater benefit as GA advances, with perhaps harm prior to 35 weeks, no real difference between around 35 and 36 weeks, and improved survival and improved outcomes among survivors born at term. Here are the Forest plots of the results for death, and for major disability among survivors, from the latest version of the Cochrane review, which dates from 2013. CoolCap, n.neuro, TOBY, Shankaran and the NICHD trial all included babies from 36 weeks, the others were limited to at least 37 weeks GA.

That Cochrane review does not include data from more recent trials, which are largely in Lower and Middle Income countries. Those trials have been reviewed several times recently, including in a very new publication which appeared in a Supplement to “Neonatology” dealing with neonatal priorities in LMICs. You may already be aware of the HELIX trial, which really threw a spanner into the cooling machine works, it was a very important multicentre trial, from India, Bangladesh and Sri Lanka, which showed an increase in mortality with cooling. As an accompanying editorial clarifies, many of the infants were SGA, were outborn and were not transported by ambulance, many had MRI injury suggesting more chronic asphyxia. Nevertheless, overall, it looks from a complete SR/meta-analysis like there is no major benefit of Therapeutic Hypothermia in LMICs, but several single centre trials have often shown a benefit. I haven’t studied this data enough to be sure about the right approach, but many neonatologists working in LMICs do still offer hypothermia, if the conditions are right.

The reason for that little detour was just to demonstrate that hypothermia isn’t always beneficial, other studies have shown increases in mortality, as also suggested by the new late preterm trial. My conclusion is that for late preterms, there is no evidence of benefit; as gestational age increases it looks more and more likely that cooling decreases mortality and decreases disability. The exact point at which the probable harm among the preterm changes to the probable benefit at term is uncertain. It probably isn’t exactly at 36 weeks 0 days. It may well depend on other factors than purely GA, such as whether the infant is SGA, the duration of the asphyxial insult, and avoidance of excessive hypothermia, as well as perhaps the severity of the encephalopathy and the occurrence of seizures. Without knowing the impacts of those factors, it is impossible currently to select 35 week GA babies with HIE who are more likely to benefit.

I am trying to imagine what I would do tomorrow if I was asked to see a baby of 35 weeks and 5 days, who was, otherwise, a perfect candidate for cooling, with a good weight, a normal fetal heart rate trace until there was a clear sentinel event like a cord prolapse, who developed moderately severe encephalopathy and seizures at 3 hours of age. I think it is too easy to just say that it would have been OK to cool if he had been born 2 days later. On the other hand, without some sort of protocolised limits then we could cool everyone! I think, right now, I would say there is no clear evidence of benefit, and maintaining strict normothermia is the best, evidence-based approach. But I couldn’t be too angry at a colleague who had decided to cool such a baby.

Posted in Neonatal Research | Tagged , , , , | 4 Comments

Therapeutic Hypothermia in the Late Preterm Infant

Posted on 9 April 2025 by Keith Barrington

Current guidelines support the use of therapeutic hypothermia for term infants with hypoxic ischaemic encephalopathy. Both the Canadian and the US guidelines include infants born at 35 weeks gestation. However, the data supporting efficacy for those late preterm infants, is extremely limited, and less mature infants were excluded from all the major trials. The discussion section of this new publication reveals that the 2 RCTs which included babies of 35 weeks GA had a total n of 7.

This lack of reliable information for a group of infants who are referred with acute encephalopathy led to the performance of an RCT by the NICHD network. Faix RG, et al. Whole-Body Hypothermia for Neonatal Encephalopathy in Preterm Infants 33 to 35 Weeks’ Gestation: A Randomized Clinical Trial. JAMA Pediatr. 2025;179(4):396-406. They enrolled 168 infants born at 33 to 35 weeks gestation, with moderate to severe encephalopathy, including a depressed level of consciousness, they did not require EEG (or aEEG). The intervention group were cooled to a standard 33.50 for 72 hours, starting within 6 hours of birth, and the controls were maintained around 370.

The primary outcome was death or disability, with disability defined as severe : Bayley III cognitive score <70, GMFCS level 3-5, blindness, or deafness despite amplification; or moderate : cognitive score 70-84 or, GMFCS level 2, treated seizure disorder, or hearing loss requiring amplification. Primary analysis was a Bayesian approach, with the prior assumption being no effect of the intervention.

As you can see there was no indication of any benefit. Death was more frequent in the hypothermia group, and outcomes among survivors were just about identical, a slightly proportion with severe disability in the hypothermia group (and interestingly, almost no survivors with moderate disability, in either group, 2 vs 0). The Bayesian analysis shows there is very low probability of benefit of cooling in this population. Also, there were many of the cooled babies who overshot their target, 32 of them had a temperature recorded under 320 during the first hour of intervention.

Interestingly, the way this is presented in the table is that there is only a 13% chance of benefit on mortality with cooling; in the discussion this is stated as an 87% probability of harm from the treatment, which gives a zero chance that the treatments are equivalent! Maybe it’s just my poor understanding of Bayesian analysis. Within Bayesian analysis there are ways of evaluating different magnitudes of treatment effect, which is something that I think we should consider for the future, rather than presenting data as if the only 2 possibilities are that the outcomes are either better or worse, it would perhaps be more helpful to calculate the probabilities of a particular, clinically important impact, and the probability that mortality is about the same, using whatever limits are considered appropriate. I guess, if the probability of benefit had been about 50%, then you could say the treatments are equivalent?

On the other hand, for this intervention, as there is very low probability that cooling is beneficial, we should not cool the late preterm infant with encephalopathy. For babies at 35 weeks, these data are much stronger than those previously available, and 35 week GA infants should probably not be offered therapeutic hypothermia, and should be excluded from the next version of any guidelines.

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