#451 - On with VON (EPS 3) - Re-examining the Evidence for Erythropoiesis-Stimulating Agents in Preterm Infants

Hello friends 👋

Does the latest Cochrane evidence finally tip the balance in favor of erythropoiesis stimulating agents in preterm infants? In this VON Grand Rounds follow-up episode, Ben sits down with Dr. Roger Soll and Dr. Souvik Mitra to review the 2026 Cochrane systematic review on early ESA use in preterm newborns. Across 37 trials and over 6,000 infants, early ESAs consistently reduce the need for red blood cell transfusion. The review also confirms with high certainty that ESAs do not increase retinopathy of prematurity. The conversation covers shared decision making, drug selection, dosing, iron supplementation, and which patient populations should be prioritized.

Link to episode on youtube: https://youtu.be/P9zTVtr2o28

----

The transcript of today's episode can be found below 👇

Ben Courchia, MD (00:00)

Hello everybody, welcome back to the Incubator Podcast. We are back for our episode with VON. This is episode three of this series, where we extend the conversation from the Vermont Oxford Network Educational Sessions. Today we are following up on the recent VON Grand Rounds webinar, "Evidence to Practice: Erythropoiesis Stimulating Agents." We are incredibly fortunate to be joined by the two featured presenters from that session. Dr. Roger Soll, welcome back to the podcast.

Roger Soll (00:31)

Thank you, Ben. And if I can take a quick opportunity — thank you very much for having us on. I love the ability to follow up from these webinars. I cannot tell you how many people stopped me to thank me for this opportunity.

Ben Courchia, MD (00:44)

That's great. We are very humbled and happy to collaborate with VON. This has been a very fruitful partnership so far and we look forward to more episodes. I also want to introduce our second guest today, Dr. Souvik Mitra. Welcome back to the podcast.

Souvik Mitra (01:02)

Thanks Ben, and always a pleasure to be back.

Ben Courchia, MD (01:04)

Always a pleasure as well. During the Grand Rounds, you walked us through the clinical landscape of using ESAs — and we will use that term throughout the podcast; it stands for erythropoiesis stimulating agents. You covered the use of ESAs to prevent anemia and improve outcomes in preterm newborns. As anyone who works in the NICU knows, a one-hour webinar barely scratches the surface. Today we are taking the time to go over the data again, dive deep into the newly released 2026 Cochrane systematic review — which Souvik, you are co-authoring, congratulations — and address some of the questions that did not make it into the live Grand Rounds. We will link to that review in the description.

So let's get into it. The first question really has to do with bias and with the new data now that the 2026 Cochrane review is out. How do we anticipate this fresh data will shift some of the entrenched practices? Roger, I will let you take this one first.

Roger Soll (02:10)

You are getting right to the heart of it, Ben. We have known for a long time that ESAs will prevent the need for transfusion — a small but meaningful effect. We can debate whether different transfusion practices or other strategies can also decrease transfusion, but ESAs have a role. The question has always been: what is the downstream effect? And how do we balance that against costs, risks, and other considerations? That has been evolving, and that is why Souvik's work is so important. There is now a huge body of evidence — 37 trials, over 6,000 infants — and we can begin to explore not just the reduction in transfusion, which both families and caregivers want to see, but all the various trade-offs. That is where Souvik's new work and his thinking around what this means for guideline developers becomes so important.

Ben Courchia, MD (03:09)

Souvik, can you give us some of the highlights of the Cochrane review? There are some very important outcomes related to reduction in intraventricular hemorrhage (IVH) and necrotizing enterocolitis (NEC), but most importantly the reduction in packed red blood cell (PRBC) transfusion, which is the most common indication for ESAs.

Souvik Mitra (03:29)

Absolutely, Ben. As Roger mentioned, the previous review showed this finding and this review confirms it: when you give ESAs early, within the first eight days of life, there is a reduction in the need for transfusion. That is not in question regardless of how we look at the evidence.

Some of the other findings concern outcomes that may matter even more to patients and families — severe IVH, periventricular leukomalacia (PVL), and NEC. And the question that has lingered for a long time and has tempered the enthusiasm around reducing transfusions is the effect on retinopathy of prematurity (ROP). We wanted to tackle that question directly in this review.

Starting with ROP: our review shows with high certainty that early ESAs do not increase ROP. We took a closer look by removing studies with higher risk of bias, and the results do not change. There is no signal for harm with early ESA use with respect to ROP. I want to put that on the table right at the outset.

The more complex questions arise with some of the other clinically meaningful outcomes. On NEC: we did find a reduction, which is very interesting, because when you look at trials that assess NEC as an outcome, we rarely come across interventions that show a statistically significant reduction. That could be very meaningful if it represents a true effect. We again removed studies with potential high risk of bias, and when we look only at the highest quality studies, the reduction in NEC is no longer statistically significant — but the point estimate still favors early ESA use. So there may still be a benefit, though we are less certain about it.

On IVH and PVL: that is where it gets really interesting. When we look at all the studies together, there is a clear effect — which is remarkable, because when was the last time a neonatal clinical trial showed a reduction in IVH, PVL, and neurodevelopmental outcomes? We see trial after trial showing no difference in neurodevelopment, and here we find some signal. However, when we restrict to only the highest quality studies, all of these effects disappear. That raises real questions about the true effect on brain outcomes.

To summarize: there are possible effects on NEC reduction, though we are uncertain. It does not increase ROP. And it definitively reduces the need for transfusion.

Ben Courchia, MD (07:25)

Roger, based on what Souvik has just presented — starting with IVH — I want to invite everyone to go back to YouTube and watch the Grand Rounds recording. There is a very interesting conversation towards the end involving Dr. Sunny Juul and Dr. Robin Ohls about the effectiveness of ESAs and how quickly after initiation they have an effect. Dr. Juul and Dr. Ohls noted that ESAs should probably be started before a baby reaches the transfusion threshold, so that the effect can be established in advance. But given that most IVH occurs in the first three to seven days of life, do we actually think ESAs play a role there, or is that perhaps an incidental finding? What are your thoughts on the mechanism behind that association?

Roger Soll (08:23)

I want to take a step back, Ben, and reiterate one key point. For me, the single biggest takeaway is that we no longer have any harms on the table. No matter how Souvik parses the data — all the data, high-quality data — that concern has been eliminated. So we are really just faced with: here is a drug that has some meaningful, moderate short-term effects on transfusion and is safe.

Your question is then: do I believe it is also beneficial for the other important outcomes that might really drive the decision making? On IVH specifically, no, I do not believe it. As you mentioned, most IVH occurs in the first three to seven days — in fact, most within the first 24 to 48 hours. A biologically plausible mechanism for ESA impact in that timeframe is not there. PVL is a different story, but again, does this drug have meaningful immediate effects? Does decreasing transfusion affect NEC? We discuss that often. Does it affect other clinical outcomes? Possibly. Do ESAs have a direct effect on various organs? Also possible. But for me, none of that is what drives the decision. The fact that we have taken harms off the table means we should rethink this entirely from a cost-benefit perspective. And I would refer back to Souvik, because he has really tackled that in the real-world context and is perhaps more cautious than I am.

Ben Courchia, MD (09:57)

Souvik, during the Grand Rounds you made a strong case for the decision to begin ESAs to be a shared decision-making process with families, presenting the evidence openly. A question that came up in the chat was: given the reduction in IVH and NEC you described — with relative risks that look quite impressive — how do you find it difficult to frame this intervention as optional? For parents hearing these numbers, what is there to be optional about? Can you walk us through what shared decision-making actually looks like when the data starts to look this compelling?

Souvik Mitra (10:55)

Absolutely. And now you are getting into the weeds of guideline development — how recommendations are made at a national or local level, and how that translates to the individual patient level.

First, this represents a significant shift in practice. As I showed in the webinar, the last time our Canadian guidelines were updated was 2015, and at that time there was a clear recommendation against use. Why? Because even then we had evidence favoring reduction in transfusion, but there was this lingering concern about harm with ROP. As Roger mentioned, we have now taken that off the table. So we are starting from a base of evidence that has shifted in favor of ESAs, but with a legacy of not routinely using them.

When our group came together, we felt we needed to be practical. That is where I always return to the evidence-to-decision framework — taking the best available evidence and putting it in the context of what matters to families, but also considering the local context, the clinical context, the practice context, and what has been done historically. At the bottom of that framework are acceptability and feasibility, which play a very important role.

When we put all of that together: yes, the evidence favors use of ESAs regardless of how we slice the data. We can debate how certain or uncertain we are about specific benefits, but there is a meaningful degree of benefit for outcomes that matter to families. The key question was whether that would be acceptable to clinicians and patients in an era where this drug has not been routinely offered.

That is why we tread cautiously and made a conditional recommendation in favor of use — which keeps the door open for shared decision-making. In practice, that means sitting down with a family and explaining that based on current evidence, there appears to be benefit and we would prefer to offer this medication, but we also need to discuss the challenges that come with it. For example, this may involve injections over a prolonged period of weeks, which can be painful for the baby. This is also something that was not previously used in our unit, but we are now considering changing our practice based on emerging data. So we want to have an open and honest conversation about why it was not used before, why we are now reconsidering it as a group, and what the realistic benefits and limitations look like. Reducing transfusion could be very meaningful for some families and may itself drive the decision. But our confidence in some of the most important outcomes — like the effect on IVH — is lower, and we want families to understand that.

Roger Soll (15:27)

Souvik, if I could jump in — and Ben, you probably saw in the chat, someone wrote, "You Canadians are so cautious." So I am going to speak as the blustering American on this. The safety concerns held me back for a long time, and I think most people will want to revisit this issue in light of new evidence. I think ESAs have great appeal now. We know they decrease transfusion. They may help with some other important outcomes. At the very least, this needs to be on the list of things discussed with families — and with a favorable slant. I do not present things as "you could do this or you could do that." It is more: "We think the evidence tilts toward this." And although as Souvik wisely pointed out, the injection experience is not pleasant for the infant or the parents, you can factor that in — but I would lean toward recommending it. And although you Canadians are cautious, I would point everyone to Keith Barrington's blog, where he says: it is time.

Souvik Mitra (16:44)

Thank you, Roger.

Ben Courchia, MD (16:52)

I want to get into a little bit of the specifics of ESAs, because while it is a useful umbrella term, we are really referring to two specific medications: erythropoietin and darbepoetin. It seems that there is a substantial body of publications on both, but we are seeing more data in recent years on darbepoetin specifically when it comes to reducing anemia. Can we interchange the data? Is what is true for one also true for the other? Do you have a sense of whether the evidence is skewing toward one medication over the other?

Roger Soll (17:45)

Let me give a general thought, and then Souvik will give the detailed analysis. Meta-analysis by definition groups a class of agents used in similar populations and gives us a result. It is one way of looking at the data, and it can shift where we stand on the usefulness of an intervention, but it always has to be deconstructed at the end: which patient, which drug, at what dose, with how much iron supplementation, and so on. Your questions are exactly right, Ben. Even if you are enthusiastic about ESAs, you still have a lot of unanswered questions to work through. I will turn to Souvik on the specifics.

Souvik Mitra (18:34)

As Roger summarized it perfectly: if you look at the data from our review, we have far more data on erythropoietin than on darbepoetin. Out of just over 6,000 patients in the systematic review, fewer than 1,000 — just over 700 — were receiving darbepoetin.

What was interesting in our meta-analysis — and this is where meta-analysis, despite all its criticisms, can be genuinely helpful — is that we can examine whether there are differences in effect estimates based on which drug is used. At least for transfusion reduction, we found no difference. The point estimates are very similar regardless of which drug you use. So from a transfusion reduction standpoint, we can be fairly confident that both drugs act in a very similar way.

For the other clinical outcomes it becomes more difficult, because event rates are lower. Comparing one drug with a large number of events against another with fewer becomes challenging. So can I confidently say there is no difference between darbepoetin and erythropoietin on outcomes like NEC, bronchopulmonary dysplasia (BPD), IVH, or PVL? No. But I can be fairly confident about the equivalence for transfusion reduction.

Ben Courchia, MD (20:21)

Roger, you alluded to this earlier — we have now broken down ESAs to two medications whose effects on transfusion are roughly comparable. But as you mentioned, the studies that were reviewed used a very wide range of doses. How do we navigate that when counseling families and presenting the positive effects? What dose do we use?

Roger Soll (20:48)

Let me start with the simpler issue of darbepoetin versus erythropoietin. The point Souvik raised about painful injections is relevant here — darbepoetin has a very different injection schedule. If you accept that the two are equivalent, even acknowledging that there are fewer darbepoetin studies, darbepoetin does become more favorable from a practical standpoint simply because of the dosing schedule and the number of needle sticks the baby receives.

On the exact dose, I will hand that back to Souvik. But importantly, I think dose optimization may be a focus for future research. I am not certain that you need to be restricted to exactly what has been done in existing trials. Think back to the surfactant research — we started at 100 milligrams, thought that was more than enough, and then found that 200 milligrams and specific delivery methods refined our knowledge significantly. Concluding that we should consider ESAs does not mean the research questions are over.

Souvik Mitra (22:03)

That is exactly right, and it is honestly one of the things we struggled with during the review. We would have loved to do a subgroup analysis based on dosage, but we did not have sufficient data to group the studies meaningfully. The dosage range across trials runs from 30 to 3,000 units per kilogram — which creates a real practical problem. If someone asked me what dose I would recommend based on our review, I would go back to the largest and most trustworthy trials for each of the two drugs and use those dosages as my starting point for clinical practice.

Ben Courchia, MD (22:59)

Thank you. Time is flying, and I want to make sure we address iron supplementation before we close. There are a lot of questions in the chat about this. The American Academy of Pediatrics (AAP) and the Canadian Paediatric Society (CPS) both recommend starting iron supplementation relatively late — several weeks after birth — and yet we are now talking about initiating ESAs quite soon after birth. How do we reconcile those two things? Should we be revising when we start iron supplementation in the context of ESA use? Roger.

Roger Soll (23:43)

Yes, I do think this requires rethinking. Again, I would defer to Souvik on the details. The first step is to look at the successful trials and, as you deconstruct the data, use their dose, their approach, and their iron supplementation protocol. But the reality is we cannot say with any certainty from this data exactly what the right approach to iron supplementation is. Your point, Ben — that this needs to be rethought and treated differently from routine iron supplementation — is valid.

Souvik Mitra (24:18)

You are right, Roger. Looking back at our own CPS guideline, if I were to redo it, I would have given much more thought to the iron supplementation section when making recommendations about ESAs, because the two are somewhat at odds. Guidelines typically recommend starting iron no earlier than two weeks of age. But mechanistically, if you want ESAs to work, you need adequate iron stores. That is why if you look at the larger trials, they started iron much earlier — within the first week. By the time babies reached an enteral volume of 60 mL per kilogram per day, they were already on iron at 3 mg per kilogram per day, titrated up to 6 mg per kilogram per day once feeds reached over 100 mL per kilogram per day. That is very different from current standard practice. So any center that chooses to adopt ESAs will need to rethink their iron supplementation approach accordingly.

Ben Courchia, MD (25:37)

And many of these trials had access to intravenous iron supplementation, which makes earlier initiation much more practical — we are not talking about giving enteral ferrous sulfate in the first days of life. That addresses one logistical concern.

On the topic of monitoring: one attendee asked about the gold standard approach for monitoring iron stores. Is it ferritin? Is it reticulocyte hemoglobin? People are wondering how to ensure we are not causing iron overload once we start supplementation.

Souvik Mitra (26:12)

Going back to the trials, the common biomarker that has emerged is ferritin. And this brings up an important point — there are significant logistical considerations that come with implementing any new intervention. That is exactly why acceptability and feasibility in the evidence-to-decision framework matter so much. If a unit does not have reliable access to ferritin measurement, or if babies are too sick or NPO and cannot receive enteral supplementation early, the practical challenges become real. To answer the question directly: based on the trial data, ferritin is the common denominator.

Roger Soll (27:01)

And just to amplify that point — once you decide you want to use ESAs, that is really just the beginning of a complicated process. You still have to figure out dose, route, iron supplementation approach, and monitoring. And it is not as simple as writing an order. This is the art of quality improvement — thinking through all the processes of care. Deciding to use the drug is honestly the least of it in some ways.

Ben Courchia, MD (27:43)

That dovetails nicely into my next question. Roger, during the presentation you referenced data from the Pediatrix clinical data warehouse showing how few patients are actually receiving ESAs — around 1.4%, though perhaps closer to 15% in some analyses.

Roger Soll (28:08)

That is right.

Ben Courchia, MD (28:09)

This led many Grand Rounds attendees to say: we want to start, we want to codify this practice. One discussion point was that this would be a very good intervention to implement specifically for babies considered to be at risk. And a lot of the chat questions have asked: who exactly is that? What population benefits from ESAs? Roger, I will let you go first.

Roger Soll (28:45)

Let me start by noting that we did not discuss HIE patients at all today — that is a separate and valid question for future investigation, but not what we are addressing here. We are talking about the preterm infant. While anemia of prematurity is a concern even in somewhat larger babies — arguably up to 1,500 grams — I do not think it represents a critical problem across that whole range. So I think we come back to the infant less than 1,000 grams or less than 28 to 30 weeks gestational age.

I would also note that I am not certain how many of these trials were conducted in the context of delayed cord clamping or current restrictive transfusion practices. That said, I would cast the net broadly and say: the target population is infants at high risk of transfusion from anemia of prematurity. I would define that on a population basis as less than 30 weeks gestational age and less than 1,000 grams. I would not parse it much further than that. There is a separate conversation about whether this applies to the very smallest babies where we have limited data, but broadly speaking, I would focus on that anemia-of-prematurity population.

Souvik Mitra (30:08)

I absolutely agree. I would add that each center should look at their own local data and identify their high-risk population, because the threshold may differ by center. Some units have strong delayed cord clamping practices and restrictive transfusion thresholds, so their overall transfusion rates may be lower — and perhaps only babies under 27 weeks represent the truly high-risk group. In another center where transfusions are more frequent even in babies under 30 weeks, the target population may be broader. Local event rates and local practices should drive your own definition.

Roger Soll (30:59)

I would add one thing, Ben: ESAs should not be introduced in isolation. This should be part of a broader effort to reduce transfusions — fewer needle sticks, standardized transfusion protocols. ESAs are a component of that. Robin Ohls writes about exactly this, and I think she is right.

Ben Courchia, MD (31:23)

Perfect segue, because the last question from an audience member asks exactly that. With the widespread adoption of delayed cord clamping, minimal blood draws, and restrictive transfusion thresholds, have ESAs become obsolete? Are these other interventions doing most of the heavy lifting? Do we still have as much need for a medication to reduce transfusion? Souvik, go ahead.

Souvik Mitra (32:09)

Purely from a data perspective, our review did not have the granularity to tease those questions apart. We would have loved to do subgroup analyses based on transfusion thresholds and cord clamping practices, but that would have required 600,000 patients rather than 6,000. In the absence of that granularity, we fall back on existing knowledge: yes, all of those practices have additive effects toward the final outcome of reducing transfusions. The question each center has to ask honestly is: with all of these practices in place, are we doing as well as we could? If collectively the answer is no, and if ESAs are feasible and practical to implement, then you could consider adding them. But you have to be honest that you should not expect trial effect sizes to replicate directly in clinical practice. The real-world effectiveness can be considerably more modest. That is exactly why quality improvement projects are so important — you have to evaluate your outcomes regularly, see if you are getting better, and if not, rethink and pivot.

Roger Soll (34:06)

Souvik, if I could add — if all the effects of ESAs are mediated purely through fewer transfusions, then Ben's point has real credence and optimizing everything else becomes the priority. But there is also the possibility that ESAs have direct organ-protective effects. So I would not discount the literature from ESAs simply because it predates delayed cord clamping and minimal phlebotomy protocols. You are right, Ben, that trials from a decade or more ago were likely not incorporating those practices. But I would say the two approaches may be synergistic — one does not discount the other.

Ben Courchia, MD (34:55)

And together they represent an opportunity for a very interesting quality improvement bundle — you do not necessarily have to choose between them, you can apply them together. That is pretty much all the questions. I think we did a nice job covering what was lingering in the chat. Roger, thank you so much for taking the time to come back on the podcast to discuss ESAs — it was a great conversation. We will see you next time for episode four of the VON series, coming in the next couple of weeks. Thank you both very much.

Roger Soll (35:39)

Thank you for having us.

Souvik Mitra (35:41)

Thank you so much.