Hello Friends 👋
In this episode of The Incubator Podcast Journal Club, Ben and Daphna explore a range of insightful articles that challenge current neonatal care practices. They begin by introducing an exciting new series called "From the Heart," focusing on cardiac journal reviews. Special guests from the EBNEO team join the discussion to dive into a recent trial on oxygen use in delivery room resuscitations for preterm infants. The hosts also examine a trial that investigates the impact of lung recruitment maneuvers before surfactant administration in extremely preterm neonates, exploring whether these techniques could reduce the need for mechanical ventilation.
This episode provides in-depth analysis and a balanced discussion about the implications of these findings for everyday neonatal care. With guest appearances from Dr. Abdul Razak and Dr. Kiran Srivastava, the team also debates the latest meta-analysis on initial oxygen concentrations during neonatal resuscitation, offering fresh perspectives on potential changes in clinical practice. Whether you're interested in innovative resuscitation techniques or curious about how recent studies may shape the future of neonatal care, this episode offers thought-provoking insights that every NICU professional should hear.
----
The HIE clinical decision tool mentioned in today's episode can be found here: https://www.infantcentre.ie/predictionapp.html
----
The articles covered on today’s episode of the podcast can be found here 👇
Vento G, Ventura ML, Pastorino R, van Kaam AH, Carnielli V, Cools F, Dani C, Mosca F, Polglase G, Tagliabue P, Boni L, Cota F, Tana M, Tirone C, Aurilia C, Lio A, Costa S, D'Andrea V, Lucente M, Nigro G, Giordano L, Roma V, Villani PE, Fusco FP, Fasolato V, Colnaghi MR, Matassa PG, Vendettuoli V, Poggi C, Del Vecchio A, Petrillo F, Betta P, Mattia C, Garani G, Solinas A, Gitto E, Salvo V, Gargano G, Balestri E, Sandri F, Mescoli G, Martinelli S, Ilardi L, Ciarmoli E, Di Fabio S, Maranella E, Grassia C, Ausanio G, Rossi V, Motta A, Tina LG, Maiolo K, Nobile S, Messner H, Staffler A, Ferrero F, Stasi I, Pieragostini L, Mondello I, Haass C, Consigli C, Vedovato S, Grison A, Maffei G, Presta G, Perniola R, Vitaliti M, Re MP, De Curtis M, Cardilli V, Lago P, Tormena F, Orfeo L, Gizzi C, Massenzi L, Gazzolo D, Strozzi MCM, Bottino R, Pontiggia F, Berardi A, Guidotti I, Cacace C, Meli V, Quartulli L, Scorrano A, Casati A, Grappone L, Pillow JJ.Lancet Respir Med. 2021 Feb;9(2):159-166. doi: 10.1016/S2213-2600(20)30179-X. Epub 2020 Jul 17.PMID: 32687801 Clinical Trial.
Gallini F, De Rose DU, Iuliano R, Romeo DM, Tana M, Paladini A, Fusco FP, Nobile S, Cota F, Tirone C, Aurilia C, Lio A, Esposito A, Costa S, D'Andrea V, Ventura ML, Carnielli V, Dani C, Mosca F, Fumagalli M, Scarpelli G, Giordano L, Fasolato V, Petrillo F, Betta P, Solinas A, Gitto E, Gargano G, Mescoli G, Martinelli S, Di Fabio S, Bernardo I, Tina LG, Staffler A, Stasi I, Mondello I, Scapillati E, Vedovato S, Maffei G, Bove A, Vitaliti M, Terrin G, Lago P, Gizzi C, Strozzi C, Villani PE, Berardi A, Cacace C, Bracaglia G, Pascucci E, Cools F, Pillow JJ, Polglase G, Pastorino R, van Kaam AH, Mercuri E, Orfeo L, Vento G; IN-REC-SUR-E Study Group; Malguzzi S, Rigotti C, Cecchi A, Nigro G, Costabile CD, Roma E, Sindico P, Venafra R, Mattia C, Conversano M, Ballardini E, Manganaro A, Balestri E, Gallo C, Catenazzi P, Astori MG, Maranella E, Grassia C, Maiolo K, Castellano D, Massenzi L, Chiodin E, Gallina MR, Consigli C, Sorrentino E, Bonato S, Mancini M, Perniola R, Giannuzzo S, Tranchina E, Cardilli V, Dito L, Regoli D, Tormena F, Battajon N, Arena R, Allais B, Guidotti I, Roversi F, Meli V, Tulino V, Casati A.JAMA Netw Open. 2024 Sep 3;7(9):e2435347. doi: 10.1001/jamanetworkopen.2024.35347.PMID: 39320892 Free PMC article. Clinical Trial.
Murray AL, O'Boyle DS, Walsh BH, Murray DM.Arch Dis Child Fetal Neonatal Ed. 2024 Sep 24:fetalneonatal-2024-327366. doi: 10.1136/archdischild-2024-327366. Online ahead of print.
Rueda MS, Soghier L, Campos J, Bahar B, Bost JE, Gai J, Hamdy RF.J Perinatol. 2024 Sep 28. doi: 10.1038/s41372-024-02120-0. Online ahead of print.PMID: 39341980
Sotiropoulos JX, Oei JL, Schmölzer GM, Libesman S, Hunter KE, Williams JG, Webster AC, Vento M, Kapadia V, Rabi Y, Dekker J, Vermeulen MJ, Sundaram V, Kumar P, Kaban RK, Rohsiswatmo R, Saugstad OD, Seidler AL.JAMA Pediatr. 2024 Aug 1;178(8):774-783. doi: 10.1001/jamapediatrics.2024.1848.PMID: 38913382
Luyt K.Cochrane Database Syst Rev. 2024 Sep 24;9:ED000168. doi: 10.1002/14651858.ED000168.PMID: 39315530 No abstract available.
----
The transcript of today's episode can be found below 👇
Ben Courchia MD (00:00.903)
Hello everybody, welcome back to the Incubator Podcast. We are back this Sunday with a new episode of Journal Club. We have lots of good articles to review. Daphna, how are you?
Daphna Barbeau (00:09.174)
doing well buddy I'm doing well just glad to be back doing journal club yeah it feels like it we've had so many things going on
Ben Courchia MD (00:13.766)
I know it's been a while.
And we have a lot of things going on, nothing, Journal Club steps for no man.
Daphna Barbeau (00:23.02)
That's right. That's right. Nothing gets in the way of Journal Club.
Ben Courchia MD (00:26.181)
Nope. We're going to have more journal club episodes. think we're listening to a lot of the feedback. People really like the journal clubs. So, let me just start by saying this. We're going to start a new series of episodes on the podcast, looking at more, trying to do more journal club reviews. So this time around, Nim Goldshtrom and Adrian Bischoff are going to do, I'm not exactly sure what the frequency is, but basically they're to do journal club just for cardiac articles. So it's going to be very exciting. It's going to be a series called from the heart and it's going to be neonatal.
Daphna Barbeau (00:32.162)
Mm-hmm.
Daphna Barbeau (00:55.37)
I love it. I love it.
Ben Courchia MD (00:55.933)
Cardiology Journal Club.
So I'm kind of excited about that. All right. We also have a special EBNEO segment this week. And we'll be joined by the EBNEO team in a little bit. Yeah. So it's going to be quite fun. I think this week we're going to be joined by Abdul Razak.
Daphna Barbeau (01:02.646)
Mm-hmm.
Ben Courchia MD (01:29.793)
and Kiran Srivastava to talk about this recent article on the use of oxygen in the delivery room. But yeah, other than that, that's really it. Should we get started? Let's do it. All right. So the article that I wanted to review today, the first article I'm going to review is an article published in 2021.
Daphna Barbeau (01:45.388)
Let's do it.
Ben Courchia MD (01:58.09)
I forget, it's kind of an old article. But the reason I'm talking about this now is because there was a paper that was published in JAMA Network Open looking at long recruitment before surfactant administration and extremely preterm neonates, two-year follow-up of randomized clinical trial. Since we never really reviewed the clinical trial on the podcast, I'm like, all right, maybe we'll start there. And this sort of happened around the time the podcast was sort of beginning. And so maybe that's why it skipped and we didn't really see it.
Daphna Barbeau (02:26.04)
slipped bias.
Ben Courchia MD (02:27.097)
Yeah, that's right. So this basically, I'm going to go first on this, to talk about this trial called Lung Recruitment Before Surfactant Administration in Extremely Preterm Neonate with Respiratory Distress Syndrome, the InRecSurE randomized unblinded controlled trial. The first author actually, I'm going to have to look it up. I think it's Giovanni Vento, and this was published in The Lancet.
So it's a very interesting concept. Predictors for an unsuccessful insure approach are basically what we know. Young gestational age, small birth weight, increased severity of RDS, hemodynamic impairment. And so what the group is arguing is that optimizing and expiratory lung volume before surfactant administration could improve the success of insure. And so what they did is that they aimed to compare the application of a recruitment maneuver, which we'll talk more about in a second.
just before surfactant administration followed by rapid extubation. So basically going from something that we all know as INSUR, intubation surfactant extubation, to INRECTSUR, intubation recruitment surfactant extubation. This is done in spontaneously breathing preterm infants to find out basically whether INRECTSUR decreased the need for mechanical ventilation at 72 hours of life. So I think we need to do a little bit of a deep dive in the study design.
Sometimes the study design can be sort of overlooked a little bit because I think some are very well known or the design is pretty straightforward. But in this case, I think it's important. So the InRecSurE trial is a randomized unblinded control trial done in 35 tertiary neonatal intensive care units in Italy. The eligibility criteria were that you were born in a tertiary NICU in the participating trial. Your gestational age was between 24 and 27 and six weeks.
You had to be breathing independently with only CPAP for respiratory support, and you had to reach this criteria of CPAP failure to then be eligible to get surfactant in this format, insure versus InRecSurE. They excluded babies that were quite sick. So if you had birth asphyxia, five-minute apgar score of less than three, if you required endotracheal intubation in the delivery room already, then for insufficient respiratory drive, if you were born after birth,
Ben Courchia MD (04:47.163)
prolonged 21 days of a premature rupture of membrane, or you had major congenital anomalies, disorders of metabolism, or high drops. So what does the intervention look like? So all neonates were started on nasal CPAP initially, received one or two sustained lung inflation maneuvers, which is basically 25 centimeters of water of pressure for 10 to 15 seconds. I know some of you may be like, But that was their protocol.
The neonates were then admitted to the NICU on CPAP. They were spontaneously breathing. An intubation decision in the delivery room was made based on an RP. So they followed an RP basically, but the infants arrived in NICU on CPAP. Now, when it came time to receive surfactant, why would they give them surfactant? And basically the criteria they use, and I know there's a lot of debate in the literature regarding what is the criteria we should be using to administer surfactant. I don't think that was the point here. They basically had a criteria and they followed it.
They said they needed a fractional concentration of oxygen in inspired air, FIU2, of 30 % or greater to maintain sats between 87 to 94 % for at least 30 minutes. I know some of you are going to look at these target sets and say, that's not what I want, but fine, it doesn't matter. Just pay attention. Then they also, the other criteria was that if their clinical status deteriorated rapidly or if they developed respiratory acidosis, which was defined as a PCO2 of 65 millimeters of mercury or more and a pH less than 7.2.
So far so good. These are the criteria they use to administer surfactants. So what are these interventions? So we're going to start with the true intervention, not the control group. So the InRecSurE group. And so basically, this is quite fascinating. I think you're going to like this, Daphna. The InRecSurE group gets intubated, and they get started on high-frequency oscillatory ventilation. And using the following ventilator settings. So they start with a mean airway pressure of 8.
a frequency of 15, a delta p of 15, and an inspiration to expiration ratio of one to two. And basically the author says that the infants undergo an oxygenation guided lung recruitment procedure using stepwise increments, then decrements in mean airway pressure to recruit and stabilized collapsed alveoli using the D. Jaeger method. So basically doing this method to try to recruit as many of the alveoli up as possible. I'm going to be very humble here. I did not.
Ben Courchia MD (07:05.797)
know what the D. Yeager method was. So I looked it up. It was published some time ago. I'm going to try to tell you it was published in the Blue Journal. And it was published by Anne D. Yeager. And the time of publication was 2006. So what is this method? It is quite interesting. How do you recruit these alveoli in inpatients before the administration of surfactant?
So I'm going to quote you this maneuver from this original paper. And so they're saying that you start at 6 to 8 centimeters of water, and that's your continuous distending pressure. And it gets increased basically in stepwise fashion, all the while monitoring the transcutaneous O2 sat and the transcutaneous PCO2. And basically, you want to continue increasing the distending pressure until you see an improvement in those numbers. Then the FiO2 is reduced stepwise.
keeping the SATs within the target range. And then the recruitment procedure is stopped if oxygenation no longer improves or if the FIU2 does not exceed 25%. The corresponding continuous descending pressure called the opening pressure, that's called the opening pressure. And then next, the continuous descending pressure is reduced stepwise until the transcutaneous CO2 deteriorates. And the corresponding continuous descending pressure, that's called the closing pressure. And after a second recruitment maneuver, the optimal
continuous distending pressure, that's called CDP optimal, is set two centimeters of water above the closing CDP, the closing continuous distending pressure. The pressure amplitude is then adjusted to maintain the transcutaneous partial pressure of carbon dioxide between 5.3 and 8.0 kPa. So it's an interesting procedure. And it obviously...
raises the point that maybe we should try to recruit these babies more and not just shove surfactants before an uncollapsed alveoli. So I think that's very interesting. Now to be thorough and continue with the intervention, they said that after recruitment procedure, the InRecSurE received 200 milligrams per kilo of pro-actant alpha via closed administration system in one or two aliquots while continuing high frequency oscillatory ventilation.
Ben Courchia MD (09:23.493)
Now the control group, which is allocated to the insure group, gets intubated and immediately receives 200 milligram per kilo of surfactant without previous lung recruitment. The infants are ventilated manually. So basically they get bagged to facilitate surfactant distribution using a T-piece device with a peak inspiratory pressure of 20 to 22 and a PEEP of about five to six centimeters of water and the respiratory rate between 30 and 40 breaths per minute.
Infants with sufficient respiratory drive then are extubated within about 30 minutes of the procedure. They're placed back on CPAP, P level of about six to eight. And that's regardless of group assignment. Both groups fall in that category. Infants in both groups who met the CPAP failure criteria again during the following 24 hours could receive a second dose of surfactant. And that dose of surfactant, which is half the original dose, is given in the same fashion that they were randomized to begin with, either the InRecSurE or just in sure. The primary outcome of the study is the need for mechanical ventilation within the first 72 hours of life.
And the infants met the primary outcome if they were not extubated within 30 minutes after surfactant administration or required reintubation within 72 hours. They powered the study to detect a decrease in the need for subsequent mechanical ventilation in the first three days from 50 to 30%. They calculated that about 103 neonates had to be enrolled to have a significant difference with 80 % power at the 0.05 alpha level. So without going further into the detail, I think we have kind of a gist of what they were trying to do.
The results are as follow 218 infants were randomly assigned between 2018 and 2023. It's not 2023, I apologize. And the per protocol analysis ended up including a bit less, 212 infants. 47 % were in the in-rec sure group and 52 % were in the insure group. The baseline characteristics were relatively similar between the two groups. 35 % were having a gestational age.
30 to 35 % had a gestational age between 24 and 25 and 6, so they were quite small. 65 % approximately had a gestational age between 26 and 27 and 6. Now, this high frequency oscillatory ventilation recruitment maneuver took about 30 minutes. That's a long time to actually get this done. I was surprised by that. I thought it would be a little bit faster, but this whole stepwise...
Daphna Barbeau (11:39.781)
Especially your first time, feels like very anxiety provoking over the course of 30 minutes, but it makes sense.
Ben Courchia MD (11:43.206)
Yeah, yeah.
Yeah. Make sure the mic is closer to you. We have a hard time hearing you. So no worries. The FIU2 value at the optimal mean away pressure in the recruitment and post-recruitment in the InRecSurE infants was 28%. It's kind of nice. At that point, maybe you're like, do we need a subtractor? And it was significantly lower compared to the insure group who had an FIU2 of about 42%.
Daphna Barbeau (12:01.836)
Mm-hmm.
Hmm?
Ben Courchia MD (12:13.317)
All the infants were extubated by 30 minutes. need for mechanical ventilation within the first 72 hours of life occurred in 40 % of the patients in the InRecSurE compared to 54 % in the intro group. So 14 % difference. Yeah. The need for mechanical ventilation within the first 72 hours of life. So was lower and the number needed to treat was seven.
Daphna Barbeau (12:25.132)
Wow.
Ben Courchia MD (12:38.597)
Now, they say they observed no difference between the InRecSurE and the regular insure groups in the proportion of infants receiving different non-invasive ventilation support strategies after extubation before starting mechanical ventilation, whether that's CPAP, BiPAP, NIMV, and so on. Some other interesting secondary outcome in hospital mortality. What was interesting about that is that when they did this according to the intention to treat analysis, they found no difference. However, when they looked at the per-protocol analysis of whoever, like,
got what they were completed their part of the trial, it suggested a protective effect of the recruitment procedure on death among infants who received the study intervention, 19 % versus 33 % with a p-value of 0.02. So it's interesting that in the intention to treat it didn't really hold up, but in the protocol analysis it did. No changes in BPD or other mortalities. The number of infants requiring two doses of surfactant in the InRecSurE about 41 % and it was much lower.
than the number of babies needing repeat surfactant in the insure group. That was 52%, but that was not really statistically significant. I think some people might have questions about some of the secondary outcomes. Like what about pneumothoracies? There was no difference. 6 % in the insure group, 4 % in the InRecSurE group. The PIE rates, 7 % in the insure group, 4 % in the InRecSurE, and then pulmonary hemorrhage as well, similar. Looking at IVH, anything...
Worse than grade two, 15 % rates in the in-sure group versus 11 % in the InRecSurE group. So nothing really statistically significant, if anything, trending better towards the InRecSurE procedure. The conclusion of this preliminary article that we are reviewing today is that before an in-sure procedure decreased the need of mechanical ventilation in the first 72 hours of life, so the InRecSurE reduced the need for mechanical ventilation in the first 72 hours of life.
And that they're saying that further studies need to be ongoing because obviously, even though the numbers were kind of like 100 in each group, they're saying it would be better if it was done on a larger group of babies. And so that's the initial paper that was published in the Lancet talking about these initial outcomes. I think that was interesting.
Daphna Barbeau (14:50.326)
Mm-hmm. I mean, we're always talking about what is the best way to administer surfactants so that it's most effective, right? So, I mean, we haven't done all the studies yet. So, cool.
Ben Courchia MD (15:01.361)
That's right. So the follow-up paper that we're going to review then is the follow-up that came out like this week or last week in JAMA Network Open, looking at the two-year outcomes of this particular trial to see how these kids are doing from a neurodevelopmental standpoint. is Arthur Galini and colleagues again out of Italy. And in this case, obviously now I'm going to go through the study design a little bit faster because we just reviewed the proper trial now. So obviously these are two-year outcomes.
The primary outcome is the occurrence of death after discharge or major disability at two years corrected postnatal age. The secondary outcomes were neurodevelopmental outcomes, major disability, cerebral palsy, cognitive impairment, visual deficit, auditory deficit. They look at anthropometric measurements. They looked at recurrent respiratory infections, hospitalization, all that in the first two years. Now, the way they measured long-term outcomes was using the Griffith's mental developmental scales.
or the Bailey developmental scale for toddlers and infants third edition. And that was really depending on the experience of the participating center. In terms of how they defined neurodevelopmental impairment, major disability was defined as the presence of at least one of the following, either cerebral palsy, cognitive impairment with a developmental quotient of less than 70, visual impairment with a visual acuity of less than 6 over 60 in the better eye, and hearing impairment. So.
You can go back into this paper, look at the, if you have more questions about the neurodevelopmental follow-up, I really, like we're already 16 minutes into it. I just don't want to take the whole episode to talk about this paper. But what was interesting is that 218 preterm infants were enrolled in the original trial. Six were excluded. It left them about 212 in the original trial, you remember, 111 and 101. And the primary outcome for this neurodevelopmental study was available in only 137 children. Obviously,
excluding the babies that passed away, some lost to follow-up. The loss to follow-up was not tremendous. It was like maybe 10 kids in each group, so it was not terrible. So the long-term follow-up at two years was evaluated in a fine population of 137 children, 64 in the regular control group, the people who received insure, 73 in the InRecSurE intervention group. So in terms of the primary outcome, there was no significant difference in the occurrence of death.
Ben Courchia MD (17:20.933)
after discharge or major disability at corrected postnatal age of two years. Of the 64 children in the inshore group, 13 died or had major disability. That's about 20%. Compared to 10 out of 73 in the InRecSurE group, that's 13.7%. That was not statistically significant, but still, again, data that's trending towards favoring this recruitment procedure. In terms of secondary outcomes,
There were no significant differences in neurodevelopmental outcomes between the different groups. Table 3 is the table you might want to look at when you're reviewing this paper. There's also no significant difference in delayed neurodevelopment when we're talking about the development quotient of less than 70 between the children in either group. There's also no significant difference in borderline or delayed neurodevelopment. That's a quotient of less than 85 between the two groups. And then in terms of long-term growth and respiratory outcomes, they didn't find significant differences.
I think their authors are concluding that these results are promising, supporting the safety of this recruitment methodology and the InRecSurE technique. And that having seen no difference between the group in death, neurodevelopmental outcomes, anthropometric measurements, or recurrent respiratory infections, their findings cannot be compared with data from other cohorts yet, they say, but further multi-centered study could be helpful as usual. So I thought that was an interesting, I mean, I think this is a very novel technique. I was not familiar with this InRecSurE procedure.
Daphna Barbeau (18:45.88)
Totally agree.
Ben Courchia MD (18:47.857)
but definitely something quite cool.
Daphna Barbeau (18:51.286)
I really like that. It's also a good reminder about this kind of optimal lung recruitment strategy anyways that, you know, some centers use right out the gate for anybody that they put on high frequency and some centers don't. So I think it's interesting.
Ben Courchia MD (19:05.251)
Mm-hmm. And that's something we started to think about in our center as well, do we need to, like, there's a rush with insurers sometimes, like, let's go, let's quickly get the tube in, quickly put the surfactant, but maybe we don't. I don't know if this particular procedure of putting kids on high frequency is something that I would be excited to take on, but once you've intubated a patient, can you give them five, 10 minutes to try to recruit and get really proper inflation? I don't know. I think that opens a...
Daphna Barbeau (19:27.682)
Mm-hmm. Mm-hmm.
Ben Courchia MD (19:32.091)
box and a question that is very interesting in my opinion.
Daphna Barbeau (19:36.48)
Yeah. Yeah. And I think, I mean, this is a whole other episode, but, you know, balancing the concern there's right. What's the concern? The balancing measure, I guess, pneumothoracies, right, is what would worry people. But it's interesting. OK. You want to go, you want me to go.
Ben Courchia MD (19:49.265)
Yeah. Yeah. All right. Do want me to go next? Go ahead.
Daphna Barbeau (19:56.562)
Okay, I actually, this is, I'm kind of doing a throwback also because there's been a lot of buzz on Neo Twitter. Dr. Karen Lloyd wrote this editorial in the Cochrane database. Basically, the editorial is entitled, Antenatal Magnesium Sulfate Reduces Cerebral Palsy After Preterm Birth. Implementation into clinical practice needs to be accelerated globally to benefit preterm babies.
And really the editorial was kind of a call to action noting that we still don't have total compliance of babies receiving mag sulfate who are eligible or fetuses I should say, getting antenatal mag sulfate who are eligible. So in 2002, she quotes only 69 % of eligible women in hospitals in the notably the global Vaughan.
receive the intervention. So that's in a thousand hospitals, 70 % of which are based in the US. So of course there are hospitals in Africa, Europe, Middle East, South Asia, Southeast Asia and South America. But we know that even in the US, we don't have a hundred percent compliance. And that editorial comes just a few months after the Cochrane review.
of mag sulfate. So this was published at the end of May and actually it's been sitting in my folder and we didn't review it. But I think it's always good to review. It's a great board question. I feel like what are the benefits of mag sulfate? So magnesium sulfate for women at risk of preterm birth for neuroprotection of the fetus. And they included randomized control trials and cluster randomized control trials.
They used six studies, a total of 5,917 women, and they had included 6,759 fetuses alive at randomization. They do note that all the RCTs were conducted in quote unquote high income countries, and they included women at risk of preterm birth at less than 34 weeks gestation. Now, they may have had slightly different treatment regimens and inclusion and exclusion criteria.
Daphna Barbeau (22:22.393)
So I will get started and tell you a little bit about the outcomes. So the primary outcomes for infants and children. So they looked at these babies up to two years corrected age. Mag sulfate compared with placebo reduced cerebral palsy, a risk ratio of 0.71. The number needed to treat for additional beneficial outcomes was 60.
Mag sulfate comparable placebo also reduced death or cerebral palsy risk ratio of 0.87. Number needed to benefit was 56. Mag sulfate probably resulted in little to no difference in death, either fetal, neonatal, or later the risk ratio of 0.96. Probably resulted in little to no difference in major neurodevelopmental disability risk ratio of 1.09. Also,
little to no difference in death or major neurodevelopmental disability risk ratio 0.95. So that's up to two years corrected. So at early school age, MAG sulfate may have resulted in little to no difference in death, fetal neonatal or later. So risk ratio 0.82 may have resulted in little to no difference in cerebral palsy risk ratio 0.99, death or cerebral palsy and death or major neurodevelopmental disability.
Then they looked at secondary outcomes. Secondary outcomes for infants or children, mag sulfate probably reduced severe intraventricular hemorrhage, grade three or four, risk ratio of .76, number needed to benefit 92, and may have resulted in little to no difference in chronic lung disease or bronchopulmonary dysplasia, or an average risk ratio of .92. They did look at the outcomes for women, so this was interesting.
Magsulfate may have resulted in little or no difference in severe maternal outcomes potentially related to treatment. So death, cardiac arrest, respiratory arrest, or risk ratio of .32. However, it did probably increase maternal adverse effects severe enough to stop treatment in average risk ratio of 3.21. Magsulfate resulted in little to no difference in cesarean section and postpartum hemorrhage. And they wanted to look at breastfeeding, but they didn't.
Daphna Barbeau (24:41.492)
enough data. So basically, the available evidence indicates that magsulfate for women at risk of preterm birth for neuroprotection of the fetus compared with placebo reduces cerebral palsy and death of cerebral palsy in children up to two years corrected age and reduces severe intraventricular hemorrhage for infants. So
We're giving it, it's a good reason to understand why we're giving it. This is a very popular board question and a reminder that even some of our institutions were checking the boxes for getting mag sulfate in like as moms are being wheeled into surgery and that's really not optimal. So how can we look at our current practices and make sure moms are getting the full dose and in an adequate timeframe to be effective for babies.
Ben Courchia MD (25:36.229)
Yeah, thank you for reviewing that. Thank you for that. All right, we're going to take a quick break and come back with our EBNEO segment.
Daphna Barbeau (25:38.402)
For sure.
Ben Courchia MD (00:00.718)
Welcome back, everybody. We are here today with another exciting EB -Neo segment. Today, I am joined by two incredible physicians and EB -Neo contributors. have Dr. Kiran Srivastava from MRR Children's Hospital in Mumbai, India. Kiran, thank you for making it onto the show.
Ben Courchia MD (00:24.556)
And we have incubator veteran Abdul Razak from Monash Children's Hospital in Australia. Abdul, welcome back.
AbdulRazak (00:32.525)
Thanks, Ben. Lovely to see you.
Ben Courchia MD (00:34.668)
Lovely to see you, the meta analysis king.
AbdulRazak (00:37.901)
I have now moved away from meta -analysis, I must say.
Kiran Shrivastava (00:44.489)
So I'm going to end this presentation with a little bit of a background on the research that we're doing. I'm going give you a little bit a on the research that we're little background research that And I'm to give you a the research that And I'm going give a background And going a And
Ben Courchia MD (00:44.918)
I feel like sometimes I'm going to take a tangent here. Sorry, we're supposed to talk about the paper. But you know, like how these news network on the TV, whenever something happens, they have the one guy that they can call. So like if there is a medical problem, they call this one doctor that they have on file. So for us, the meta -analysis, have Abdul. So whenever meta -analysis comes out, it's like, let's talk to Abdul about that. But today we are talking about a very
AbdulRazak (01:01.74)
you
Bye!
Ben Courchia MD (01:10.67)
polarizing paper that came out in JAMA Pediatrics a few weeks back. The title of the paper is Initial Oxygen Concentration for the Resuscitation of Infants Born at Less Than 32 Weeks Gestation, a systematic review, an individual participant data network meta -analysis. The first author is James Sotiropoulos, who we've had on the podcast as well when the paper came out.
Kiran Shrivastava (01:20.86)
Okay.
Ben Courchia MD (01:35.618)
This paper really is a who's who of the literature regarding neonatal resuscitation. We have the likes of Ola Saugstad on there and many other important researchers, physicians, scientists, and neonatologists. And even though we may have covered this paper a few weeks back on one of the journal clubs, Kiran, I'm wondering maybe for the audience who needs a bit of a refresher, can you remind us...
Kiran Shrivastava (01:57.116)
This is a very interesting way of thinking the United States.
Ben Courchia MD (02:02.01)
what the paper is about or what is the question they were trying to answer and what did they find?
Kiran Shrivastava (02:08.188)
Sure. So this was called the Net Motion Trial, and this was a systematic review and network meta -analysis of individual participant data from which they took from all randomized control trials that have been done in preterm infants that were born at less than 32 weeks gestation, and that evaluated what initial oxygen concentration was used for their resuscitation. So they managed to get data on 1 ,055 babies from 12 of 13 eligible trials.
and they analyzed 1 ,003 of these babies. They did this by defining three different concentrations of initial oxygen, which was low, which was if they were resuscitated with starting with 30 % or less of oxygen or intermediate, was like a tight 50 to 60 % and high, which was more than 90 % oxygen. What they found was that in the primary outcome, which was all cause in hospital mortality,
This was lower in those infants that were randomized to the initially high FIO2 group compared to those that received low or intermediate oxygen. So just in terms of numbers, this looked like 9 % of about 350 babies that received high oxygen died compared to 14 % of those that received low oxygen and 15 % of the 169 babies that received intermediate oxygen.
They did not find any significant differences in other premature morbidities that they reported as secondary outcomes, which was chronic lung disease, intraventricular hemorrhage of a severe degree, or any retinopathy of prematurity. They also found that those babies that received higher initial oxygen concentrations had better odds of achieving target saturations of 80 % at five minutes of age compared to those that received low oxygen.
Ben Courchia MD (04:02.862)
Yeah, this was quite impressive. as you mentioned, I think at one point when you're talking about mortality, the number needed to treat that was mentioned was 16. And I think that's quite impressive. Can you guys tell us a little bit about the methodology? Because I think that we are accustomed to seeing systematic reviews. We're accustomed to seeing mid -analysis. But this did something a little bit unique.
Kiran Shrivastava (04:13.948)
16.
Ben Courchia MD (04:27.822)
It's not really unique, but it's something that we don't see every day in our literature specifically, which is individual participant data network meta -analysis, which means that, well, I guess you'll tell us what this means.
AbdulRazak (04:34.049)
Thank
Kiran Shrivastava (04:40.313)
Abdul, do you want to take that?
Ben Courchia MD (04:40.535)
Abdul, you want to take this?
AbdulRazak (04:41.579)
Yeah, yeah, that's fine. Yeah, think they, I mean, everybody knows meta -analysis nowadays, I guess it's not nothing new, but this one is relatively different than what we normally do. Normally in meta -analysis, they take a couple of RCTs and pool them as long as they are homogeneous enough to pool them. But in this one, what they did was they actually took the raw data. So it is like capturing individual patient data from every single trial. And then,
Kiran Shrivastava (04:44.444)
you
Kiran Shrivastava (04:59.963)
Thanks for watching.
AbdulRazak (05:10.797)
pulling all those data into the meta -analysis. That's a little different than how we traditionally do. Now, the major difference that we get is how we define the outcomes, how we define the interventions. So many trials define the outcomes in different ways, such as mortality, some people say before discharge, some people say 36 weeks, some people say two years. But here in this individual participant data meta -analyst,
Kiran Shrivastava (05:13.616)
because we're not going to We're going to be to that. We're going to that. We're going to be to that. We're going to that.
Kiran Shrivastava (05:40.016)
Thank you.
AbdulRazak (05:40.095)
the authors generally do is from the beginning, they define what the outcome is going to be defined. And they just pull out all the raw data. So basically they're asking, us all the hard work that you have done in all those years. So that's an important data that people share. Then once that's given, and then they pull all the data into the study, and then they pull all studies together.
Kiran Shrivastava (05:56.285)
I'm out. I'm out. I'm out.
AbdulRazak (06:05.107)
So that's the endial part, the spendidum talents. here, other than the meta -analysis itself, they just didn't do the usual meta -analysis. They also did the network meta -analysis. So normally in meta -analysis, we generally have two arms. One is the intervention, and another one is the control arm. But here, we see there are three arms. So one, we have low oxygen, high oxygen, and intermediate oxygen. So when we have more than two groups, then you can cross -compare each other, which is not possible.
Kiran Shrivastava (06:17.087)
Thank you.
Ben Courchia MD (06:26.69)
Mm
Kiran Shrivastava (06:30.17)
Thank you.
AbdulRazak (06:32.671)
in the traditional meta -analysis. Traditional meta -analysis compares the ones which are comparable. But in the network, if A is compared with B and B is compared with C, you can compare A versus C. So that's how you do the network meta -analysis. And that's what they have done. So it's a very unique and very unique combination of what they have done. So it's individual participant data plus network meta -analysis. It's a quite robust.
Kiran Shrivastava (06:42.562)
Thank
Ben Courchia MD (06:48.546)
Mm -hmm.
Kiran Shrivastava (06:56.444)
This is a very similar thought. But it's a very similar thought. So I'm going to stop. This is a very similar thought. So I'm stop.
AbdulRazak (06:58.573)
Take what they have done and that speaks to the credibility of the data itself So whatever data we see it's more granular and more authentic than the traditional pairwise meta analysis in traditional pairwise What happens we are a bit worried about how people define but here so we are more happy how they have defined this more homogeneity So there's no more belief in what we are seeing but again, it depends on the confidence or the
Kiran Shrivastava (07:15.099)
you
AbdulRazak (07:25.879)
certainly what comes up, but this is the basic difference between the network and usual meta analysis.
Kiran Shrivastava (07:30.724)
So, thank you.
Ben Courchia MD (07:31.446)
Yeah, I mean, this is quite unique, obviously, as you mentioned. And it also shows how much work was needed to actually perform this particular analysis. And I think, as you said, it's a testament to the collaborators who all shared the raw data. Many people are reluctant sometimes to do that for whatever reason. people sharing the data, but also for the primary investigators of this paper to actually go through that vast number of patients to actually
Kiran Shrivastava (07:35.816)
And
Ben Courchia MD (08:00.642)
provide the analysis that they published. I'm going to go back to you, Abdul. mean, in terms of the study itself, at first glance, it seems to be very robust. It seems that the results are quite reliable. And interestingly enough, after every paper, especially in neonatology, it seems like there's always a little
AbdulRazak (08:10.082)
Yeah.
Ben Courchia MD (08:27.47)
disclaimer of like, more is needed. But when you read this paper, you're like, maybe not. Maybe that's pretty definitive. So can you tell us a little bit about the limitations of this paper and how you're interpreting the results themselves?
AbdulRazak (08:44.299)
Yeah, I think when we speak to limitations itself, the limitations of individual participant data, the major limitation that comes up with the number of trials that get dropped off. Normally what happens when we approach the authors, many would not agree. They say, is my data I'm not going to share with you. And that's how you lose the data. And when you lose the data, then you actually don't have confidence in it because you have not pulled all the data. So that's the problem with that. But here,
Kiran Shrivastava (08:58.235)
you
Ben Courchia MD (09:01.612)
Right.
Ben Courchia MD (09:10.733)
Mmm.
AbdulRazak (09:13.261)
they were able to approach all the authors and they were able to pull all the RCTs data. In fact, I think they pulled up 12 out of 13 RCTs. I think one RCT they tried asking them but they never replied back. But that RCT was very minimal patients and I think that's not a major issue at all. So technically if you ask about the study itself limitations, there are none, to be honest. But when we come to speak about the limitations of the evidence, that's when...
Kiran Shrivastava (09:28.516)
you
Ben Courchia MD (09:36.387)
Mmm.
AbdulRazak (09:41.901)
you know, we pick up our notes and say, yes, there are many, then you can really think whether we should rely on this or not. So study itself, I would say there are no limitations. So I would say it's incredible job what they have done. But then I think you really pulled out the point about the mortality that's in fact startled me as well when I looked at this paper. So if you look at the mortality in the low group, was 50, 14%, I guess.
Ben Courchia MD (09:53.037)
int
AbdulRazak (10:10.637)
And if you look at the mortality in a high group, it is 9%. So there you go. There's a 5 % difference in mortality. Can you believe it? The 5 % difference in mortality in a mutated paper. that too, just giving some. Yeah, it's just very little difference in the intervention. You're just giving 21 % oxygen versus 60 % or 100%, whatever it is, for only a few minutes. Because after a couple of minutes, most of the babies are on 100%.
Ben Courchia MD (10:22.226)
Where do I sign?
Kiran Shrivastava (10:22.47)
Thank
AbdulRazak (10:37.677)
You know, we are so worried that we increase the oxygen to 100 percent. I think here, you know, just looking at the intervention, it's just lasting shortly, but how much impact it could do. So that's a big thing that we should think about. That difference is huge. The number two, I would say before going to limitations itself, I would say that this is not a new meta analysis question that people have not been asking. We have been asking this question for past two decades.
Kiran Shrivastava (10:41.552)
I'm going to show you how to make a good, good, good, good, good, good, good,
Kiran Shrivastava (10:59.46)
you
Ben Courchia MD (11:06.872)
Mm -hmm. Yeah.
AbdulRazak (11:07.437)
I think we started asking this in 2008 or 2005 and now UN until 2023 papers are going on and they have been many systematically meta -analysis that have already been published looking at all these RCTs. I think if I'm not wrong, there are at least four of them. There might be many, but at least to my knowledge, there four of them. And all this meta -analysis Ben and Ketan, as you see, they have showed no difference. And the question is now this meta -analysis has really shown difference. That's a big...
Kiran Shrivastava (11:10.736)
Thank you.
Ben Courchia MD (11:31.448)
Mm
AbdulRazak (11:35.405)
Why did it show a difference? Why the other meta -analysis could not show the difference? If you closely look, what previous meta -analysis, what previous authors have done is basically combine the intermediate or to intermediate to high. So they basically said, okay, less than 50 % is low and more than 50 % is high. So they just came up with this arbitrary number and that is probably discussed with the LCR committee and say like, look, it's less than 50, I'll go with low, more than 50, I'll go with high.
Kiran Shrivastava (11:54.216)
Thank
Ben Courchia MD (11:58.765)
Yeah.
AbdulRazak (12:04.781)
So what happens with this is the intermediate group, which is 50 to 65, get pulled in the high group.
Ben Courchia MD (12:11.534)
Yeah, and that's a key thing that I would like to mention because I'm sure some people were like, wait, what were the groups again? Like they divided it, as you said before, in three groups, low, intermediate, and high. The low was basically 30 % or less, meaning you were giving resuscitation with 30 % or less. The intermediate, as you just mentioned, was 50 to 65. And then the high percentage group was more than 90.
AbdulRazak (12:22.059)
Yeah, in this.
AbdulRazak (12:30.187)
Yep.
Kiran Shrivastava (12:32.376)
More than 90.
AbdulRazak (12:32.653)
more than 19. Yeah, but as we discussed the previous meta -analysis, they said less than 50, I guess, was the ball point. Below than that is low and above than that is high. So in that way, what has happened is the intermediate group got pulled into high group. And because the intermediate group had high mortality, now the high group also went up on high mortality. That is why the previous meta -analysis could not show any difference. But now with this one,
Ben Courchia MD (12:42.082)
Mm -hmm.
Ben Courchia MD (12:49.133)
Mm
Kiran Shrivastava (12:58.662)
Thank
Ben Courchia MD (12:58.711)
Mm -hmm.
AbdulRazak (13:00.577)
kind of getting into more nitty -gritty and more granularity saying, OK, yes, let's just do it. And then they start to find difference. That's a huge difference. that particularly, we think about why. And that simple intervention only lasting for a couple of minutes could make such a big impact. That's a big difference. That's the, yeah, sorry, go ahead.
Ben Courchia MD (13:20.685)
Yeah, I -
No, no, I was going to say I wanted to talk a little bit about that because I think that right now the big question everybody is asking is, well, should we put this into practice? Should we wait for formal recommendations to be coming down the pike from Ilcor or from the NRP in the US? But and what's interesting to me is that I was talking to one of my colleagues, Dr. Mitch Stern, about this because we are there's a few of us in our group who are instructors for resuscitation. And we were reflecting on this that
Kiran Shrivastava (13:38.233)
And point. I think that's a good point. I think that's a good I that's good point. good point. I that's a I think a good point. I think that's a good point. I think that's a good think I think a good a point. I
Ben Courchia MD (13:50.446)
A very common mistake by trainees who are learning the process is that they tend to forget to turn on the oxygen. They go through the steps of going up on the pressure or they may go up, they do suctioning, they may even think of advanced stairways and they forgot to turn on the oxygen to increase the FIO2, which is surprising, obviously, because it's such an easy intervention. But it is something people tend to sometimes not remember. And maybe this approach might solve this by saying, well,
we're going to remove this and you're going to start at a high level of oxygen and then make your way back down. Kiran, I'm going go to you on this one. Do you think that the data is there to actually start shifting practice a little bit in the delivery room?
Kiran Shrivastava (14:33.916)
So like Abdul said, the study had a lot of, so the study was done actually in a period which spanned 2005 to 2019. And I think in the neonatology world that has probably been errors of change in how we receive and resuscitate our preterm babies in that period. Along with that, there was a lot of heterogeneity in what protocols were followed.
Most of the studies, think about 10 of them were done in developed nations in properly resourced tertiary NICUs, but two of them were done in low and middle income countries. And obviously the protocols varied. They also found a lot of difference in how the oxygen was titrated even after they chose one oxygen to start with.
Ben Courchia MD (15:16.184)
Hmm.
Ben Courchia MD (15:26.305)
I see.
Kiran Shrivastava (15:26.704)
They wanted to actually analyze the effect of a fast versus a low titration effect, but they were not able to do that because there was so much inconsistency in the data. And also just going back to which places these were done at. So one of the studies was, I think the one from, the one that was done in India didn't actually have an oxygen blender, which lets you decide how much oxygen by how much you are going to go up or down on the oxygen.
Ben Courchia MD (15:37.709)
Mm.
Kiran Shrivastava (15:54.492)
So their titration was just starting at room air and giving oxygen, which would be 100 % oxygen. So there was no way of ascertaining how much oxygen or how fast it went and what the effect of that was. So I think given all of that, this study just sheds light that we need much more studies and much more data in this area, looking at the different variables that affect how these babies.
oxygenate and how we can support them into transitioning better. And I think there are a few studies that are doing this or are planning to do this. And I think it would probably be prudent to just wait and see what these studies show.
Ben Courchia MD (16:38.348)
Wow, Kiran is very high standards, very high expectations. No, but it's interesting. It's interesting you remind that I'm curious to hear your thoughts then on the difference that you brought up between I don't like the term developed country, but countries with high resources versus low to middle level of resources. Do you think that the adoption curve should be different depending on the amount of resources is available to?
the providers and the different families and stuff.
Kiran Shrivastava (17:10.096)
I think so, because there definitely is a difference in what is available in terms of resources and also in terms of training and education in these places, which I think also doesn't just depend on where you're even in a properly resourced country, there may be differences in the training of who is resuscitating these babies. but I think the policymakers and the recommendations need to consider this going forward.
Ben Courchia MD (17:37.388)
Mm -hmm. Mm -hmm.
AbdulRazak (17:38.495)
And Kiran, just to add one more point, we have a very different population than as you know, I think in developed countries, in countries which are developing, we have a lot of SGA babies and they react differently. The pulmonary results particularly react differently to those. So that number is going to be, I mean, the effect will be going to be different. So if a study is going to show a difference in a developed country, might be something not the same in the country which is developing. So we have to be careful of.
Ben Courchia MD (17:41.038)
Peace.
Ben Courchia MD (17:49.879)
Yeah.
AbdulRazak (18:07.263)
The population differences is the important difference that would suggest apart from the training and other things as well.
Ben Courchia MD (18:07.32)
So.
Ben Courchia MD (18:14.03)
So as we try to wrap up this conversation and let's try to put like, as we said, let's try to put a bow on this and try to see what we can leave the audience with. It seems to me that what I'm gathering from your commentary and from your input is that until now we were very fearful of turning up the oxygen because we said, my God, I'm going to cause all these things. If I go above 50%, if I could really resuscitate this baby with 30%, it would be better. But like you said, I think look at every patient individually, make your decision. But if you need to,
Kiran Shrivastava (18:23.608)
I'm sorry, I'm sorry.
Kiran Shrivastava (18:43.606)
And
Ben Courchia MD (18:44.108)
wrap up the oxygen for a specific patient where you think this might be beneficial in achieving stabilization in a more timely manner, then you shouldn't feel this fear that maybe was there before this paper came out. And maybe you should feel that you have some support in your decision to go ahead with that potential intervention. Do you think that that's okay to leave people with that?
AbdulRazak (19:05.773)
I don't know what to leave it to them to be honest. Yeah, I think this paper definitely I would say many neurologists now would be more rushing to turning up the oxygen. I mean, I don't know what the L -Core is going to come up, but most likely they would say stick with what they have done before. The main reason for this would be
Kiran Shrivastava (19:25.562)
Thank you.
AbdulRazak (19:30.655)
say the low certainty of evidence in the confidence estimates that we've got. I think that's a big thing that I would say even Ilcor might come up with the same thing. But I would say with this paper, many neontologists would try to crank up the oxygen little faster, a little quicker. Everybody does different ways, but I think people are trying to get into good oxygen. That's going to definitely happen. There's no doubt on it.
Ben Courchia MD (19:54.712)
Yeah, But I'm not saying that we should rush to do it. What I'm saying is in the case where you feel this might be beneficial, you shouldn't feel the inertia that maybe we felt before of saying, that's probably like we, some of us may have thought this is really bad if I turn this oxygen up a bit too fast, but now you can say, well, there's some data, low certainty, as you said, that might show that I'm not potentially going to cause more ROP. I might actually improve mortality. Let me just, let me just try this. And like you said, every patient is going to need to be treated individually. And we're not, we're
Kiran Shrivastava (20:04.537)
because
Ben Courchia MD (20:25.11)
What I gathered from you guys is that the evidence is definitely not strong enough to make a blanket recommendation, say like everyone should go through this. But there's going to be these patients where this might not be such a bad idea if you need to go that route. Kiran, any parting thoughts for us?
Kiran Shrivastava (20:39.471)
No, I think I agree. The ILCO recommendation even now still says starting at 30%, but then titrating to target saturations. I don't think that approach changes. So if you need to help a baby meet.
your target saturation, think you need to go up on the oxygen if you need to. But also remembering that obviously the oxidative stress that comes with high oxygen needs to be balanced. And I think this is why we need more data to see which way is the best way to help these babies.
Ben Courchia MD (20:59.789)
Mm -hmm.
Ben Courchia MD (21:08.726)
Yeah. And in an ideal world, if you could resuscitate an infant with the lowest amount of oxygen possible, then that's the ideal scenario. I think we should not lose sight of that. What were you going to say, Abdul?
AbdulRazak (21:10.123)
Yeah, I got it.
AbdulRazak (21:21.385)
Yeah, I think that's right Ben. I think the one question I wanted to ask you is, let's imagine you have this population data of thousand babies with so much of heterogeneity and there's one more trial which is coming up with 1500 babies data which is a lot of homogeneity. What are you going to do with it? Are you going to wait or are you going to wait, use the evidence at this point and try cranking up the oxygen?
Ben Courchia MD (21:42.166)
Yeah, I think that's the problem that the L -Core faces because they have to make a recommendation that is applicable to pretty much everyone. And that's tricky because I can look at my specific population in my division where we know exactly the types of babies that we have. They're pretty consistent across year over year. And we can make a decision to say, well, for a specific kind of infants, maybe some of our more immature infants or some of the babies that really have a hard time achieving target oxygen saturations.
Kiran Shrivastava (21:43.804)
.
AbdulRazak (21:48.62)
Yeah.
Kiran Shrivastava (21:54.628)
because we are not going be do that. We are to do that. Yes. We are going to be able to do that. We going be able to do that. Yes.
Ben Courchia MD (22:10.402)
because of surfactant deficiency and so on, maybe we can look at being a little bit more liberal, but that may not be applicable to everybody. And so I think, like you said, look at your patients, look at what you're doing, but consider now that this is a potential option. think that's what I'm getting away with. Abdul, Kiran, thank you so much. We've extended our time, but it's always a pleasure talking to you. And thank you to the EB -Neo team for...
Kiran Shrivastava (22:34.036)
So, in terms of how this is being done, we are going to have a very complex, very complicated, very complicated that is really going a challenge for the students.
Ben Courchia MD (22:38.54)
always providing these compelling commentaries and staying on top of the evidence. You can find out more about the eBneo team and the eBneo commentaries on eBneo .org. And you can actually volunteer to write up commentaries like the one that you guys wrote. You'll be paired up with an eBneo mentor and you'll get this published in Acta Pediatrica and you can get the opportunity to come on the podcast to chat about a paper. So it's a very collaborative process that's a lot of fun as I, as I'm from what I'm hearing, right, Kieran?
Kiran Shrivastava (23:08.272)
Yes, it is definitely. Thank you Ben.
Ben Courchia MD (23:09.39)
Thank you both and have a good rest of your day.
AbdulRazak (23:10.644)
Thank you
Ben Courchia MD (25:45.327)
Okay, I think that that was great. Thank you to Abdul and Kiran for participating. More EBNEO stuff coming down the pike as well. And an interesting article about the management that really has made some waves. I think you're going to like the next article that I have ready for us today, Dafna.
Daphna Barbeau (25:48.535)
Mm-hmm.
Daphna Barbeau (25:58.872)
for sure.
Daphna Barbeau (26:05.327)
Let's see.
Ben Courchia MD (26:06.981)
It's called validation of a machine learning algorithm for identifying infant at risk of hypoxic ischemic encephalopathy in a large unseen data set. First author is Anne L. Murray. It's in the archive of the Eases in Childhood, Fetal and Neonatal Edition. It's study out of Ireland.
Daphna Barbeau (26:14.968)
Hmm.
Daphna Barbeau (26:22.122)
I am interested.
Ben Courchia MD (26:23.557)
Oof, you don't even know the half of it. So the reason I wanted to review this paper is because I think sometimes we talk about machine learning, we talk about AI, but it feels very distant to us. So the aim of the study was to validate the previously reported machine learning algorithms to identify infants at risk of HIE immediately after birth using real world readily available clinical data.
We're going to play some games. You're going to enjoy this. I'm going to first go over the study design. So this is a retrospective review of electronic health record data from all-term deliveries in Cork University Maternity Hospital during a five-year period from January 1, 2017 to December 2021. Inclusion criteria all infants above 36 weeks who had a blood gas drawn within one hour at birth. Inclusion criteria that basically if you were less than 36 weeks, if you were missing data or
Daphna Barbeau (26:57.258)
Okay.
Ben Courchia MD (27:23.557)
Again, if you were less than 37, six weeks, six weeks because of the consensus not to cool patients below this gestational age. So there's a previously trained open source logarithmic regression and random forest plot prediction algorithms that is available now online. So this tool that we're going to talk about today, you can all go and play around with it. We'll do that together right now.
it's actually accessible. And basically what they did is that they used this tool that's available on a website to calculate the probability index for each infant for the occurrence of HIV. And these models were trained and assessed on a separate data set, right? So the data set that produced the model that's available on this website is the B-HIVE2 cohort of infants with sign of perinatal asphyxia with and without HIV matched against healthy cohort. And so
So now this algorithm is going to be asked to predict the risk of HIE on this data set from Cork University Maternity Hospital, but has never really been trained on that. So it's going to see how it performs basically. The algorithm require the five variables that involve an ABGAR score at one minute, an ABGAR score at five minutes, postnatal pH, a base deficit, and a lactate. And that's
That's the stuff that you input into the computer, and it will generate for you a probability index, depending on which model you use, either the LR, the logistic regression model, or the random forest model. And it will give you a probability going from 0 to 1, with higher values indicating a risk of HIE. So I can tell you what their data shows, or you want to play around with it. You want us to...
What do you want to do first? Should we run a few simulations? All right. So let me just open the app. OK, so definitely let's play around. Give me the Abgar score at one in five minutes of a baby. Yeah, everything hypothetical, obviously. And we'll play around. So.
Daphna Barbeau (29:21.527)
Yeah, let's do it.
Daphna Barbeau (29:31.041)
Okay.
of a hypothetical baby.
Daphna Barbeau (29:40.106)
Okay, let's say three and six.
Ben Courchia MD (29:42.329)
Okay, so let's start. So I've got to score at three. At one minute it's three and then at five minutes it's six. What would you give as you kind of need a lactate level? we have to give a lot. have to give... Yeah, you have to guess. But I mean, I think we... Okay, we're to put six. What about your bass deficit?
Daphna Barbeau (29:52.153)
But I have to guess okay, let's say six
Daphna Barbeau (30:01.226)
let's say minus 15.
Ben Courchia MD (30:04.073)
minus 15. And what about your pH?
I mean, if you're minus 15, you're probably like 6, 9, 7, 7, 0. OK, fine. We're going to say 7. So according to the model, yeah, the logistic regression model gives you a probability of 80 % of having HIE in this scenario. The random forest model gives you a probability of 94%.
Daphna Barbeau (30:08.83)
We're probably at a yeah we're probably at a seven.
Daphna Barbeau (30:24.918)
Mm-hmm.
Daphna Barbeau (30:29.994)
Okay, let's do a more mild baby.
Ben Courchia MD (30:30.813)
Let's do it. All right, let's go. Abgar scores, one minute, five minutes.
Daphna Barbeau (30:35.904)
Let's say four and seven. Four and eight. Let's say four and eight.
Ben Courchia MD (30:40.253)
4 and 7, 4 and... Okay, what's your lactate now? 2, okay. What about the base deficit? Minus 9. And what is your pH?
Daphna Barbeau (30:46.057)
two.
minus nine.
Daphna Barbeau (30:53.59)
Let's say 7.08.
Ben Courchia MD (30:58.077)
1.5. OK, 7.1. OK, in that case, the logistic regression tells me that the probability of HIE is 30 % and the random forest is 15%. I love, I mean, Daphne is going to ask me for this. I know, but it's an interesting decision tool. So let's go back to our paper.
Daphna Barbeau (31:08.524)
Mm-hmm.
Hmm
Daphna Barbeau (31:15.423)
But they still don't tell me what to do.
Daphna Barbeau (31:21.27)
Yeah. It's funny when you think about it in your mind though, if you were to get those sets of that set of information, wouldn't that be your guess on how much the likelihood of each?
Ben Courchia MD (31:30.215)
Mm-hmm.
Ben Courchia MD (31:34.705)
Well, that's the whole study, basically. So the study was done in a way where they're going to submit five years worth of patient data to this algorithm and see what it predicts. Now, including in this data set, the investigators know which babies had HIE because they have the clinical course available to them. So it'll be interesting to see. And they have MRI data for some of these babies, right? So.
Daphna Barbeau (31:36.472)
Mm-hmm.
Daphna Barbeau (31:51.565)
Mm-hmm.
Sure.
Daphna Barbeau (31:59.138)
data.
Ben Courchia MD (32:01.821)
So let's see what happens. So over the five-year study period, 1,191 infants above 36 weeks had a blood gas measurement. Of these, 1,081 had a complete data set available for them to use. 7 % had a diagnosis of HIE. 93 % did not have HIE. So let's see how the model fares with that. Of the 76 infants with HIE, 37 were mild, 29 were moderate, and 10 were considered to be very severe.
60 % of the HIE group received therapeutic hypothermia, 100 % of the kids who were considered severe, 100 % of the kids who were considered moderate, and 19 % of the patients who were considered to be mild, showing again this therapeutic creep that I think we've all seen. The primary diagnosis for non-HIE patients, so the other group of patients, it was usually your typical NICU stuff. 44 % had respiratory issues, 23 % had asphyxia but without encephalopathy, and 17 % had suspected infection, and then there was obviously
the rest, which I haven't calculated. But it's basically miscellaneous stuff, rashes and all sorts of other weird things. So let's look at how these two models did. First, let's start with the random forest. So the random forest had an AUROC for the prediction of HIE that was 0.926. So quite good with a p-value of less than 0.001. The median probability index in the HIE group was 70 % compared to 0.05.
in the non-HIE group. So it discriminated pretty well. With the random forest model using a probability index cutoff of 0.3 to optimize sensitivity, 936 of the 1,081 infants, which is like 86.5 % of the infants, were correctly classified by the algorithm. Sensitivity being 86.8, specificity being 86.6. Now, of the infants with HIE, 86.8 were correctly identified.
and 13.4 % were incorrectly classified as HIE.
Ben Courchia MD (34:05.597)
Now looking at the logistic regression model, the AUROC is 0.928, so very similar to the random forest. The median probability index was 0.71 for the HIE group, 0.05 for the non-HIE group. The sensitivity and the specificity using a cutoff of 0.3 was 84 % for the sensitivity, 88.7 % for the specificity.
84.2 % were correctly identified. Only 11.3 % were incorrectly classified as HIE when they did not have HIE. The random forest model, they say, using the cutoff threshold of 0.3, identified 78 % of mild HIE, 93 % of moderate, and 100 % of severe HIE cases. Some interesting data here. Of the eight cases of mild HIE that were incorrectly classified as non-HIE,
None of the babies developed seizures and non-received therapeutic hypothermia. I thought that was interesting. They also say that in the two cases of moderate HIE that were incorrectly classified, one had clinical and electrographic seizures and also had an abnormal MRI. And the other did not. And so there's table two that basically will look at the different elements of HIE by severity. But the conclusion of this paper, because I don't want to dwell too much on it because we're running out of time, is that in a large unseen data set,
An open source algorithm could identify infants at risk of HIE in the immediate postnatal period when an infant had a probability index score of 0.3 or more. This should prompt a focused clinical examination and may aid in the identification of infants who require transfer to tertiary care and timely neuroprotective intervention. like in the...
In the second patient you gave us, your random forest probability was 0.15, which may say like you're fine. The logistic regression was just 0.3. So do with that what you wish. But the other one clearly told you that you had to be concerned about the more severe presentation. So we'll link to this prediction app that's online. It's at infantcenter.ie. Center spelled the British way C-E-N-T-R-E. So infant.
Daphna Barbeau (35:56.908)
Mm-hmm.
Ben Courchia MD (36:22.413)
center.ie but we'll put the link on the episode show notes so you can play around with that. So
Daphna Barbeau (36:26.902)
And the truth is, mean, all teams studying HIE are looking at this, right? Like, how can we better predict which kids will benefit from cooling and which kids will not? For sure, for sure. Very cool.
Ben Courchia MD (36:32.027)
Yeah.
Ben Courchia MD (36:36.262)
Mm-hmm.
Ben Courchia MD (36:40.129)
Yeah, I have one more small paper if you want me to, can I go ahead? Okay. So this was a paper in the Journal of Perinatology called Blood Volume Collected for Cultures in Infants with Suspected Neonatal Sepsis. First author is Maria Rueda. my God, we're going to have fun on this one as well. So blood culture is the gold standard for diagnosis of neonatal sepsis. We all know that it has high sensitivity to detect bacteremia in concentrations of one to 10 colony forming units per ml when sufficient blood volume is collected.
Daphna Barbeau (36:44.896)
Yeah, let's do it.
Daphna Barbeau (36:54.772)
a great paper. Love it.
Ben Courchia MD (37:09.489)
Now, apparently, there's not just the AAP that has guidelines on how much blood we need from blood culture, but the Infectious Disease Society of America. They have a consensus recommendation for the blood culture.
Daphna Barbeau (37:20.225)
Pretty reputable, you know organization gosh, gosh
Ben Courchia MD (37:23.005)
careful because I'm about to tell you how much blood they want from you. But I mean, it's a yeah. So the blood culture volume collection that they recommend is weight based. And so they're saying that you need a minimum of two ml if the baby is less than a kilo or four ml if the baby is under two kilos.
Daphna Barbeau (37:41.853)
So the nurses have stopped listening at this point.
Ben Courchia MD (37:44.241)
The AAP says that at least one ml per bottle for all preterm inferences in 34 weeks undergoing substance evaluation. We're going to go with that. Am I right?
Daphna Barbeau (37:48.002)
Yeah.
Daphna Barbeau (37:52.066)
But you know, I mean, we think about this, but you don't just get one culture sometimes, a lot of times you get, you're getting two cultures a lot of times. It's a lot of blood. It is a lot of blood for these little babies.
Ben Courchia MD (38:11.527)
Sorry, I was muted. Yeah, absolutely, absolutely. And so we're going to talk about that in this paper. So the study really aims to determine the blood culture sample volumes collected by weighing inoculated blood culture bottles. We'll talk about that as well. Identify clinical environmental factors associated with lower recommended volumes and compare the sample volumes of patients treated for culture negative sepsis with patients who had actual sepsis. So this was a both prospective and retrospective study. they did it in two.
Daphna Barbeau (38:13.57)
That's okay.
Ben Courchia MD (38:38.237)
two different epochs, whatever. And it happened at Children's National Hospital in Washington, DC. It's a big 66-bed unit. It's a referral center. And so how did they find out how much blood was collected? So the inoculated volume is calculated by subtracting the weight of the un-inoculated bottle from the measured weight of the inoculated bottle. So basically, weight the difference and dividing the weight in grams by 1.06 gram per ml to account for the density of the blood.
And so basically the way this works is that after the blood culture collection by the neuro provider, all the inoculated cultures bottles were weighed after the arrival to the microbiology lab by the technologist and the weight in gram were documented. They were documented. Then they also determined the weight of the un inoculated bottles. And they did that on a, on 100 unused blood culture bottles, which seems like a wasteful because without the disposable cap.
and they get the average weight and then they confirm that number by validating this data point by weighing an additional 50 unused aerobic and anaerobic blood culture. yeah.
Daphna Barbeau (39:43.402)
It seems like it would be easier to just have the nurse document how much they drew.
Ben Courchia MD (39:48.135)
we're going to talk about that. So anyway, so what they consider to be a lower than recommended volume was if you had 0.9, less than 0.9 ml or less, the bloodstream infection was defined as an isolation of organism that's in the bloodstream from two or more blood specimens collected on separate occasion. Additionally, as defined by the NICHD,
When presented with a single cons positive blood culture, a bloodstream infection is considered if the patient has a CRP that's above one within two days from the blood culture or if it was treated by the clinician for more than five days. A contaminant is considered if you have a common commensal organism, if it's identified on a single specimen for single positive cultures with cons, those who do not meet the criteria mentioned above for the bloodstream infection would be considered contaminant. Culture negative sepsis is a patient that has no identifiable source of infection.
in the context of a negative culture that's treated with antibiotics for more than 72 hours. And then the sepsis rule out patients that in the context of a negative blood culture were treated with antibiotics for less than 72 hours. And they use the cutoff of 72 hours to be able to include all patients with antibiotics who were discontinued for a short period of time. So the results, 742 blood cultures taken from 219 NICU patients between 2018 and 2019.
The median volume of blood inoculated into blood cultures was 1 ml, and the mean was 1.06. And they found that only 35 % had less than the 1 ml recommended. So most of the time, the team does a good job.
Daphna Barbeau (41:22.914)
Well, because you don't want to be told to have to do it again. That's the worst.
Ben Courchia MD (41:26.934)
So let's dig a little bit into. So what I loved about this paper is that they not only looked at the number that the number of bottles that were inoculated with less blood, but they also looked at when did that happen? So in the univariate analysis, blood cultures obtained during the night shift were associated with the collection of lower than recommended sample volume. This difference remained in the multivariate analysis and no significant.
Daphna Barbeau (41:51.715)
The rebellious night shift.
Ben Courchia MD (41:55.613)
No significant differences in blood culture volume were noted by patient gender, chronological age, weight at the time of culture collection, or the source of the sample. So it seems that at night we consider ourselves satisfied with maybe subpar volumes in the blood culture. A few more results of the 742 blood cultures identified with documented volume. 20 % were from patients treated for culture negative sepsis. 14 % had confirmed.
bloodstream infection, 39 % underwent a sepsis rule out and 26 % were diagnosed with another infectious diagnosis. The mean inoculated volume was 1.06 ml for culture negative sepsis, 1.03 ml for sepsis rule out and 1.09 for bloodstream infection, which were not significantly different from one another. The proportion of blood cultures with inoculum less than 0.9 ml was 36 % for sepsis rule out, 32 % for culture negative sepsis and 30 % for bloodstream infection.
Again, staying around this one-third, one-fourth type of number. I'm going to stop here. The conclusion are that blood culture inoculum in their NICU was generally consistent with the guideline recommendations. Collection of samples during the night shift was associated with lower than recommended inoculum volumes. Quality improvement efforts are needed to increase the proportion of samples with adequate volume. The volume of blood sampled does not differ in patients with culture negative sepsis, bloodstream infection, sepsis rule out, suggesting that this should not be a justification for longer duration of antibiotic therapy.
Interesting. Interesting.
Daphna Barbeau (43:22.86)
All right, very interesting. Very cool.
Ben Courchia MD (43:27.215)
All right. All right. This was, do you have anything else for us, or should we wrap up?
Daphna Barbeau (43:32.824)
think we got to wrap it up today.
Ben Courchia MD (43:34.874)
Let's do it. Thank you for taking part in Journal Club today and thank you for the eBneo team for participating and we'll be back this week with more episode and another interview this coming Sunday. Daphne, thank you so much.
Daphna Barbeau (43:48.044)
Sounds good, bye everybody.
Ben Courchia MD (43:49.553)
Bye, everybody.
Comments