In this episode of the Society of Critical Care Medicine (SCCM) Podcast, host Marilyn N. Bulloch, PharmD, BCPS, FCCM, welcomes Matheus Bannach, MD, of Urgency Hospital of Goyes in the Eruseo-Dorne Cancer Hospital in Brazil. They discuss Dr. Bannach’s article, “Transfusion Practices in Traumatic Brain Injury: A Systematic Review and Meta-Analysis of Randomized Controlled Trials,” published in the April 2025 issue of Critical Care Medicine.
The study compared liberal and restrictive strategies for patients with traumatic brain injury (TBI). They found that a liberal transfusion strategy results in better neurologic outcomes than a restrictive strategy. Drs. Bulloch and Bannach discuss the importance of optimizing transfusion thresholds, the risks and benefits of blood transfusion for critically ill patients, and the limited supply of donated blood.
The conversation also covers study methodology. Dr. Bannach explains the choice of main end point for the study, the process of selecting articles to include, and the rigorous peer review process.
Listeners will find guidance for optimizing blood transfusion in patients with TBI, as well as key insights into review methodology. More discussion of this article can be found in the April 2025 Critical Content video.
Resources referenced in this episode:
Transfusion Practices in Traumatic Brain Injury: A Systematic Review and Meta-Analysis of Randomized Controlled Trials (Larcipretti ALL, et al. Crit Care Med. 2025;53:e963-e972).
Dr. Bulloch: Hello and welcome to the Society of Critical Care Medicine podcast. I'm your host Marilyn Bulloch. Today I'm joined by Dr. Matheus de Andrade Bannach, MD, to discuss the article Transfusion Practices in Traumatic Brain Injury, a Systemic Review and Meta-Analysis of Randomized Controlled Trials published in the April 2025 issue of Critical Care Medicine. To access the full article, visit ccmjournal.org. Dr. Bannach is a neurosurgeon with a fellowship in neuro-oncology and practices at UGA or the Urgency Hospital of Goyes in the Eruseo-Dorne Cancer Hospital. Welcome Dr. Bannach. Before we start, do you have any disclosures to report?
Dr. Bannach: Hello. It's a great pleasure to be here and to discuss this article. And no, I have no disclosure.
It's just a topic that I researched for research purpose. It's not really anything that would compromise my analysis.
Dr. Bulloch: Okay. Well, great. Well, I enjoyed reading your article.
I found it to be really, really interesting. In the background of your article, you mentioned that the main hypothesis for why patients may have increasing morbidity could be related to the development of anemia, which in turn delays healing because there's this impaired oxygen delivery. And that leads to poor neurologic outcomes.
From a practical consideration for our listeners, what poor neurologic outcomes do you personally see the most in this situation?
Dr. Bannach: This question actually, it's well discussed in the literature before because we always had this hunch that you need to protect the surrounding areas around the main trauma. So this is actually something that we try to alleviate the secondary damages. So everything we do in the trauma setting after a major trauma is to protect the surrounding area.
So this notion that you need to deliver oxygen to this region is actually the main focus of the patient care in the neurocritical ICU or wherever this patient is admitted at. So this notion, it depends on where the major trauma is. When we're talking about high intensity traumas, usually related to car or automobiles accidents, it's more common here in my country.
We majorly see frontal and temporal lobe contusions. So this contusion leads to altered state of mind. Usually they get really confused and memory is quickly impaired.
Memory, we know it's spread out through the brain, but it usually is really damaged by this diffuse traumas. And sometimes you see behavioral changes with aggressiveness and sometimes seizures. That's the main thing that I'm usually seeing the trauma patients.
Dr. Bulloch: So it seems like these are kind of a little bit varied, but still I think that would be scary for any patient to have. I want to talk just a moment about the methods of your study, because this was a systemic review and a meta-analysis, right? Yeah.
Now after your literature search, and then when you pulled out all the duplicates, you used something called the RAINN software for systemic reviews to sort of help you screen titles and abstract. I know some of our listeners are probably really familiar with this software, but for those our listeners who maybe have never heard of this software before, can you describe how it works? Is it easy to use?
Is it expensive?
Dr. Bannach: It's actually really common. Usually we use it to just assemble together the articles and then you can more easily go through the names and the abstracts, because if you do that on the platform, usually it's not really practical. So this software assists us in doing that.
I believe there's a free version of it, but you can add features with the paid version. There are many types of softwares that you can use. This is just a practical one that we are used to use, but it's very straightforward.
There's nothing really major that is done there.
Dr. Bulloch: There's just something that caught my eye with all the discussion nowadays about AI and machine learning and other software things to help structure workload. It seemed a lot more efficient than traditional hands sorting through all of the titles and abstracts.
Dr. Bannach: Yeah. Nowadays, because you have many types of platforms that you have like NBase and PubMed and Cyhub, each one has its own workflow. So it's usually hard to do the same thing methodologically, consistently through all of them.
And the chance of being duplicates are higher. So you need something to work through these things and help you to exclude duplicates and assess the many abstracts for your meta-analysis. Because if you don't do that, you might discover later that you're talking about the same articles or the same population, or you leave something out.
So a software that can aggregate every study found in every platform and automatically exclude duplicates, it's usually something that accelerates the process of doing the meta-analysis.
Dr. Bulloch: Interesting. Let's talk about the studies that you did include. Now, your eligibility criteria included the requirement to compare different hemoglobin or hematocrit thresholds for red blood cell transfusion, which you emphasized as being either liberal versus restrictive.
Now, all of the studies that you included defined restrictive as a hemoglobin of less than seven grams per deciliter, but at least two of the studies listed a liberal threshold differently than the others. Do you agree with the labeling? Do you feel like clinically, if you had to put threshold on liberal versus restrictive, you would have done the same thing?
Dr. Bannach: This is a thing that confuses us. And actually, for me, these labels are, at first sight, it seems counterintuitive. Actually, it should be the other way around.
If you're going to be more restricted, you should have something that secures and diminish the indications for transfusion. And liberal seems more like you're going to transfuse more. And at first sight, because the restrictive is with a lower hemoglobin level, it seems like this should be restricted, but this should be the liberal one.
But actually, it's because the quantity of transfusion that you're going to do with a higher hemoglobin level, it would be more on the liberal strategy. Because if you need to maintain an hemoglobin level higher, you will need more transfusion. So you will be more liberal with the indication.
And if you need to maintain a hemoglobin level lower, you will be more restrictive with the indications for transfusion. So that's the logic behind it. But I agree with you.
That's so confusing. It appears at first sight that should be the other way around.
Dr. Bulloch: It does. And it's interesting because so much in medicine, we get these terms that are never really given a consensus definition, but someone uses them. And it's almost like it, by default, becomes the definition, even though we don't know how it got defined that way.
And that certainly came out when we saw that two of the studies didn't conform to the way the rest of them were.
Dr. Bannach: That's one thing about meta-analysis that usually makes it harder, because sometimes different authors will refer to the same thing with different names. And other times, they will refer to different things with the same name. So it's challenging sometimes to aggregate every study that is talking about the same thing and doing the same study.
And this sometimes is challenging for us that are doing meta-analysis and review to make sure that we're not including different things. And we usually say that you're not comparing apples and oranges. And at the same time, you're not excluding things that are slightly different, but they are actually talking about the same thing.
Dr. Bulloch: Absolutely. So it can be especially confusing for some of our learners that are just trying to get up to the medicine lingo. Now, another thing I noticed in your exclusion criteria was that you excluded studies that had, quote, overlapping populations.
And if that happened, you chose the study with a larger population. Is this something that actually occurred when you were narrowing things down? And if it did, could you maybe provide some examples where it occurred?
Dr. Bannach: No, actually, it didn't happen. But this is a standard criteria of exclusion when you're doing a meta-analysis, because it's very common for authors to expand on an initial study published prior. So sometimes the authors publish a smaller trial with a smaller sample, but then they expand and publish a larger one with more patients.
So when you're doing the meta-analysis, you can include both because you'll basically be doubling some of the patients. So you need to exclude the lesser one and include the bigger one because that way you include every patient that was actually treated and you have the results for each and every one of them. And you're not doing something that duplicates some of the results, because especially if it's a significant study, you can reinforce the conclusion of this study if you include both because you're doubling down the patients.
You need to exclude these patients.
Dr. Bulloch: But that didn't happen here. No duplicate studies, no double up. I just wanted to make sure.
Now, your study's objective was to evaluate transfusion practices in patients with brain injuries and specifically evaluating neurologic outcomes. And to that point, the study's main endpoint was a favorable Glasgow outcome scale or an extended GOS at six months, which is a metric that's used to assess the patient's functional outcome after a TBI. Was there an opportunity in any of the studies to evaluate maybe a more clinically based neurologic outcome?
Not, I know that this is a standard, but did any of the studies look at things that patient may not know what a GOS is, but they may know if their memory is better, you know, they perform ADLs or something along those lines?
Dr. Bannach: That's the trouble about neurological studies that assess the functional outcome because neurological outcomes are a very diverse term. This is not something that is straightforward. You can go straight to functionality and even functionality is something that will differ between patients because for one patient, for example, for an engineer, the capability of doing math is more important than for a chef, for example.
So they actually, how to define what functionality is, it's troubling for us. It's a simpler scale that usually focuses on the capability to communicate and to deambulate, which leads to functional outcome in terms of being an individual that can take care of itself. That's the main goal of preserving neurological functions.
It's for the individual to remain independent and not reliant on some external help. So when you're doing that, usually you leave out these more peculiar things, like these details of memory, of whether you're able to recognize your family members or if you can do math. Some of the studies nowadays try to include a neuropsychological examination, but it's a long formula and you have to compare different domains and individualize the responses so it's a complex thing to just assess the entirety of a neurological outcome.
So the GOS is a simple way to assess why either this patient is independent or it's dependent for the basic activities of the daily life. This is the main goal of any neurological intensity treatment. It's to preserve that independency.
So when you're talking about GOS, usually a bad outcome is a patient that is dependent, that's reliant on external help for daily activities. And when you're looking at a good outcome is a patient that is independent, that can take care of itself, like take a shower, brush your teeth, and eat. That's the difference that is assessed in most studies.
Dr. Bulloch: So it seems like the GOS is a really good surrogate, though, of these potential clinical outcomes. Is that what I'm hearing?
Dr. Bannach: Yeah, it's simple, because that's the problem. If you get too complex things, you have the problem of heterogeneity. If I start to assess many things, for one combination of impairments, this person is independent, and for another patient, they are dependent, but they actually have almost the same level of neurological performance in the daily life.
So when you're talking about a randomized controlled trial, you need to diminish the probability of heterogeneity that will transform the analysis in a complex thing that you have to individualize too much of your sample. In this sense, the GOS is a very good scale to assess daily life activity functionality. So that's why we use it.
Nowadays, we have more advanced analysis, a modified GOS that includes more subtypes of outcomes. So we're increasingly trying to better this assessment, but the GOS actually is a good scale for this daily life independence. This is a very good scale.
Dr. Bulloch: Very good scale. Now I want to move into the results for a little bit, because I think the process of systemic review is so interesting. I feel like we in medicine have gotten almost programmed to expect large randomized controlled trials and evidence-based medicine, and that sometimes, especially for younger clinicians, it can be challenging when that evidence doesn't exist.
Now your literature search initially yielded 3,913 results, which is a lot. Once you actually applied all of your inclusion and your exclusion criteria, you only had five studies that really met your criteria and could be included in the systemic review. Was this amount what your group expected to find when it started the process?
Dr. Bannach: Yeah, that's due to a lot of things, but it's fairly common in the meta-analysis process. You find an initial result of more than 2,000, sometimes 5,000 papers, because you need as general as you can probably do your initial research due to the chance you might exclude something because you didn't use the correct term on your search. So you need a more broad term search to include every possible study and with every possible variation of the keywords that you're typing in.
This helps you to not exclude something because you didn't find it. And this in the meta-analysis process is very important. Of course, you can't use too broad term because all the way you get like 10,000, 50,000, 100,000 articles and that it's humanly challenging for us to completely assess the inclusion criteria on each and every one of this.
So usually we're trying to reduce in about 2,000 to 10,000 papers on initial research. And then you can apply your inclusion criteria on them and find the ones that really matter for your design question. So it's very common.
The initial search you do by title because on the platforms, generally, the keywords are too broad. So you get a lot of things that is not related to your research question. So the initial exclusion, it's very pragmatic.
You like read the name of the article and then you just exclude it. Then you get a smaller sample and you filter out the ones that you want based on their abstracts. And then you select the ones that are going to full reading.
And then on the full reading, you conclude the application of your inclusion criteria to select only the ones that are using your extreme parameters of inclusion and have the complete data and have like a good methodological process of doing like they did the methodologically right study and the quality of the data is good. So on meta-analysis, it's extremely common. You just have to record each and every step of this because so many people get confused on how to do a meta-analysis and you just have to do it in a way that is reproductive because one of the problems that we have nowadays in science is reproducibility of the trials.
This is a question that is being raised again and again, especially due to the larger and larger number of studies being published every day. So you have to be really careful of doing a thorough process of elimination and using extreme criteria to do that in a way that is reproductive.
Dr. Bulloch: Was your team expecting to only find five? Did you look at the results when it was done and be like, man, I thought there'd be a little bit more than that?
Dr. Bannach: Because it's an important question and something that is discussed again and again and again, I would actually expect some more studies, but also because it's not an interventional study and you need larger and we only wanted a randomized controlled trial, it's really hard to do a large multi-center randomized controlled trial on something that there is no financial gain for in the industry or something that will incite researchers to look and do these larger trials. So when you're talking about something that is like trauma and there is no really specific intervention, you're just doing transfusion, it's not something that you can patent or there's not a specific drug, it's really hard to find randomized controlled trial, large, multi-centered.
So despite expecting more trials, I'm not surprised that there are so little.
Dr. Bulloch: It kind of really reinforces the value of your paper, you kind of bring all that data together. Now let's talk about what you found. Your primary endpoint was that Glasgow outcome scale and you didn't find a significant difference in favorable JOS and really none of the studies showed a significant difference.
In fact, only one study, I believe it was Bob Robertson and colleagues, even trended towards favoring the restrictive approach. Now based on what you've read in the literature and your own clinical experience, is this surprising given the results that we know occurs in non-TBI patients? What do you think about that?
Dr. Bannach: Yeah, like I said, this is a question that it's always troubled us because when you're talking about prevention of secondary lesions in TBI patients, you have like many things that are done to ensure that the surrounding area is getting oxygen and glycols at sufficient levels. So it's like intuitive that you want more hemoglobin to be available to transport oxygen and you have more efficient delivery of oxygen and in this way you preserve more this surrounding area. So this question always intrigued us because despite this physiology and our understanding of how the brain injury evolves through time, we didn't ever find some significant results of maintaining this hemoglobin level higher, which is what inspired us to do in our meta-analysis was the publication of the EMOTION trial, which is a large trial of transfusion and these exactly things that we just discussed. So this trial, you can read through it and you see that the trial aims to prove that there is actually a significance because everything they assessed was just to try to minimize any intervention and to find this thing that was never proven, that's the better neurological outcome, but they didn't find it. So this intrigued us very much and actually our initial analysis was that there was no significance in transfusion for maintaining these hemoglobin levels, but during our peer review, we actually had a publication of another trial, so we had to redo all the analysis to include this larger trial, which is the TURGEN trial.
Dr. Bulloch: That's the TRAIN trial you mentioned to me offline.
Dr. Bannach: Yeah, the TRAIN trial. And the TRAIN trial actually used a different threshold. We already had this thing before.
We had the GOBATO trial, which he used a different threshold, which is 9, prior to redoing the meta-analysis. And this trial was the only trial that actually reached significance. However, this was a small trial with only 60 patients, so we didn't have enough of evidence based to say that the 9 threshold would be beneficial.
But this is our initial analysis that said, maybe there's something here, maybe this alter actually found something, because like I said, it's our basic instinct to think if you keep a higher hemoglobin, you will have better oxygenation. So physiologically, it's expected to reach better neurological outcomes, but nothing seemed to work in the trials. So initially, we discussed that on our initial assessment.
Maybe that's the thing that should be studied. Maybe that's the thing that's missing. We have to change the liberal strategy threshold to 9, and maybe that will result in better neurological outcomes.
But the review process actually criticized us because we were doing that. Our colleagues found that we couldn't say that. But to our surprise, during the peer review process, we had a publication of the strain trial, which is a large trial, multi-centered, randomized, and all the things that we talked, and then we had to redo.
And curiously enough, this trial used the 9 threshold also, and it also found the statistical difference. So to our surprise, actually, we did find when we included that, but it took us redoing everything, and leaving one study out, and dealing with the heterogeneity, and doing all the math again to ensure that we're not like forcing a result to make sure that we're not seeing what we want to see.
Dr. Bulloch: I found it interesting that you said that you had to go back and may have changed your conclusions, and I think that made things click for me, because when I was reading your discussion, you had mentioned that your finding showed better neurologic outcomes with the liberal transfusion group without increased mortality. I was like, wait a second, I thought they didn't find any significant difference. But it sounds like that was kind of written with the perspective of these two trials that had that higher threshold.
Is that right?
Dr. Bannach: Yeah. When you include only the trials with the 10 threshold, you find no neurological outcome difference. But when you use only the two that does 9, you do.
And then you put them all together, you find that there is actually a significant better neurological outcome in using the liberal strategy. And that was really the main focus of our conclusion. If you have better outcomes with this 9 threshold, and now we do have a meta-analysis that is our trial that says that with a larger trial, like the trained trial, that says that 9 is the threshold, like maybe we should revise our guidelines and use that threshold as a starting point to transfusion.
One of the things that may justify this difference is because the brain has a delicate balance between you can't have too much blood going into it, because you have the closed box relation. I don't know if most people are familiar with that thing, but your brain is actually a closed system. It controls everything that goes in and out and is actually able to control the blood flow by its own.
It doesn't need your heart to do it. Actually, your heart responds to your brain in terms of maintaining the ideal amount of blood going in and out of the brain. So when you're talking about a trauma patient, you have to be really delicate about controlling the parameters that you need, because if you enable too much blood to go into, you lead to brain edema.
It's bad for a recovering brain. However, if you let too little blood go into, you lead to ischemia. So you have to balance this out.
That may be the problem. When you're giving a too much blood transfusion, you might lead to edema, and you lose out the benefit of having more hemoglobin. While if you use a threshold a little lower, you achieve that hemoglobin level that actually gives more oxygen, while not leading to too much blood going to the brain.
So maybe that's the explanation to having these differences in between the two thresholds of liberal strategy.
Dr. Bulloch: Makes sense, and it certainly would be a great opportunity for someone to explore in future research. Now, let's talk about some of your other endpoints that you didn't find any difference at all in terms of mortality, whether it was hospital follow-up or ICU mortality. I think the endpoint that doesn't surprise anybody at all is in the liberal group, they got significantly more units of packed red blood cells, ranging from a mean of 3.1 to 7.1 more units per patient. Now, we're looking at this from a resource conservation perspective, because what it's like in Brazil, but here it can be difficult. You can have blood in very short supply. Using the data that you've provided, can we use that to maybe help conserve these types of precious resources?
What do you think?
Dr. Bannach: That's another thing about our conclusion, because despite our intuition that you should transform, you couldn't do that in the reality settings, because there is no proof that this leads to a better outcome. So that's the main problem surrounding this question in the brain trauma studies. Like, how do you justify giving more and more blood if there is nothing that proves that this leads to a better outcome?
And you're right about the scarcity of blood. And not only that, but transfusion leads to complications. You can have the host problem, like you have allergic antibody reactions and you may lead to systemic reaction that it's actually bad for the patient.
So it should rely on good evidence that you are improving something so we can use it. And the neurological outcome is this good reason to do it. Because when you're talking about using this scarce and actually really expensive thing, because blood is really expensive.
I don't know if many people know that, especially patients usually don't know that. But despite blood being donated by patients, usually they are charged to the health system. So whenever you're indicating transfusion to a patient, that leads to a high cost to the health system.
So you need to have this good evidence-based decision that you're doing the best thing. So when thinking about the 10 threshold, you not only don't have the benefit, but you're actually wasting on this precious research. And the 9 threshold seems to be the exact amount where you are going to transfuse more, but also you lead to better outcome and you don't transfuse as much as the 10 threshold.
Because as I explained before, the higher you need the threshold to be, more transfusion you're gonna require to do so. Because when you're talking about a trauma patient, usually they evolve with anemia because of the trauma, because they have bleeding, because they usually are patients with multiple trauma. So they had bleeding and they have a systemic reaction to the trauma that leads to reduction of the production of red cells.
And they have these multiple things that are leading to anemia. So you need to constantly be repopulating the red blood cell to maintain that hemoglobin level higher. So the higher you want it, more blood is gonna be transfused.
So when you have a lower, like a 9 threshold, you can transfuse less blood. And in this sense, the 9 threshold is better because you transfuse less and you get the benefit. So the best thing for now, I believe after this trial, is to maintain above 9, the hemoglobin level.
Dr. Bulloch: You had mentioned in there that giving blood is not without consequences. And there's certainly some things that have popped out in your study where the liberal group had the highest prevalence of ARDS. And I know that your discussion mentioned that transfusions can lead to hypervolemia and acute pulmonary edema.
And you had a few other thoughts that you outline in your discussions. You know, what do you feel is the most likely factor of factors that led to this increased risk?
Dr. Bannach: Yeah, the problem is blood is not completely compatible. That's the thing. Even if you completely cross-examine the bloods, and we only test the major causes of a host reaction, like AB and RH negative or positive.
So these are only two of the components that we have So when you're giving blood, despite being the same type and the same RH negative or positive, there are many other molecules in that red blood cell that can lead to an inflammatory response on the host. So when you're talking about transfusions, there is no completely compatible blood that you can transfuse. So what we do nowadays is that we use the best matched blood that we can.
But even though we do that, there is going to be some inflammatory response to that red blood cell. And this inflammatory response is what leads to the other lesions that we are talking about. So if you have that inflammatory response in your lungs, you're going to have lung problems.
And if you have that on your blood vessels, you're going to have another type of response to transfusion. And usually we see mild reactions like fever, sometimes spiking in C-protein levels, and you might see some increase in the parameters of renal function. But depending on how incompatible are these other factors besides AB or RH, this could lead to severe consequences.
So when you transfuse more, you're actually using more types of blood because you're not going to get like 5 or 10 transfusions from the same donor to that patient. You're going to have to use multiple sources of blood. So the more you need to transfuse, the more risk you're bringing to that patient that his body will react to these other factors that are in the blood that might lead to a complication.
Dr. Bulloch: Do you think there's anything we can do with the bedside maybe to reduce the risk or you think it's just something that we need to accept and monitor?
Dr. Bannach: In bedside, you need to monitor it really carefully. You can't ask for donation from blood relatives. This generally reduces the amount of compounds that are incompatible in this setting.
So if you have your brother or your mother or your sister donating blood to you, if it is AB-compatible and RH-compatible, they will have less of these other things that can lead to an inflammatory response. At bedside, you can test every relative that you have available and ask them to donate if that's possible. And on the long side, there is nowadays research going on for development of artificial blood, which could result in having less of these parameters of incompatibility.
So we have less reactions. We also advise having a protocol to doing the blood transfusion. So you need to prepare the patient, sometimes using steroids to reduce the inflammatory response to that inflammation, and also being really careful to control everything that is going on with the patient in order to not letting the hemoglobin levels go lower due to some other infections.
Like you have to control bleeding, you have to control the inflammation, and you have to expand the volume and controlling the renal function and everything that could save you some transfusions and diminish this volume of external blood needed.
Dr. Bulloch: Artificial blood under development, I think that's a podcast for another day, but it'll be, I think, really interesting in several years down the road, maybe look at the results from using that versus using real blood. But like I said, that's a discussion for another day. I have one more outcome I want to talk about, and that's the instance of thromboembolic events.
There really wasn't overall a significant difference, but there was one study, and that was by Robertson and colleagues, that did actually show more of these events with the liberal strategy. Why do you think this particular study's results were so different from the other two studies that were included in evaluating those events?
Dr. Bannach: It's an older study, so possibly this leads to the evolved knowledge on the area and maybe controlling more these things, because despite it's not general knowledge, and even if you are a physician, you're not completely aware of every combability test that is done to make sure that you can do that transfusion to that patient. So maybe there's something to do with being more careful in the transfusion with time. When you're talking about smaller studies, it's really hard to point out what exactly went that on the methodology or maybe the one thing that led to this.
It might be just chance. You can't really say. There was the criticism about our initial assessment that nine might be the threshold, because it could relay only on chance.
If you have a specific sample of patient that is not completely representative of the larger population, you could have a finding that is not actually consistent when you go to larger and larger populations, because it might be some characteristics of that specific sample that you're looking at. And the only thing that made us really change our mind and change the reviewer's mind about affirming the conclusion that we had was the publication of the larger trial with the same thing and consistent. So proven that this is something that is consistent when you increase and you go to larger samples.
So maybe that's that, but there's no way of assessing what justified that.
Dr. Bulloch: Interesting. And I think we're getting to the end of our time here today. This is a really interesting conversation and helpful to a lot of people in critical care, especially those that are in the neurocritical care space.
But before we leave, is there anything we didn't get a chance to talk to about your paper or transfusions and TBIs that you want to discuss for a little bit?
Dr. Bannach: When we did the analysis, because we have a constraint of physical space on the journal to publish, you have to leave out some calculations and some analysis. And I believe that if we further exploit this data, we might even have some more conclusions about that. One thing that was discussed on the peer review process was the effect model that we use.
Because when you're talking about different studies, you have to think about how the heterogeneity might play a role in the results. So you have to decide whether you use a fixed model base or an aleatory effect model. This is going too deep.
Maybe the audience isn't accustomed to these terms. But what we're saying is just that you need to say if you expect the studies to be too different or too similar. Because if you're dealing with especially similar articles, you may consider that they are controlling the parameters in the same way.
So you can use the fixed model. And when you're talking about studies that have too much heterogeneity, you might use an aleatory effect model, which is more conservatory in terms of less reliant on the conclusions that you're getting, because you're expecting the heterogeneity to play a larger role. So initially, we did our analysis using an aleatory model because of the samples that were used, because of these divergent thresholds that were used, different population sizes, and everything that led to clinical and threshold heterogeneity.
But the actual calculation of heterogeneity was low. It was lower than 40%, which is the threshold that you use to see if something has a low or moderate heterogeneity. So it was advised to us to use the fixed model.
And when you do the assessment on the aleatory model, you get no significance. And when you do it on the fixed model, you get the significant result. So at first glance, you might say, so you're using the only thing that proved what you're saying.
If you used the different model, it would not be significant. But when you think about it, the one thing that is adding up heterogeneity is the different thresholds. And if you do subgroup analysis and isolate the groups, like doing thresholds of 10 and threshold of 9, you get a much higher effect at 9 than when you include everyone.
And so the increase in the heterogeneity led to a dispersion of the effect. I don't know if I'm getting too deep right now. And if I am, please say so I can explain in a more simple way.
But what I'm saying is if I used the high heterogeneity model, it would actually worsen my results due to the inclusion of the heterogeneity led by the higher threshold. So this fact actually contributes to the notion that the 9 threshold actually is the best one to use because if you isolate that factor, you get the significance because you have the two trials that use it and both of them found their best neurological outcome in transfusion. While if you include only the 10s, you get no benefit.
And if you use the aleatory model, despite not being significant, you have a great tendency in towards the significance, which means that this thing actually contributes to the notion that we should really use the 9 threshold as the transfusion threshold for now.
Dr. Bulloch: That's a great place, I think, to leave our episode on with a recommendation from Dr. Bannach to use that 9 as the official threshold, at least for now until another study comes along and changes our mind or confirms it for sure, as happens all the time in critical care. Dr. Bannach, thank you so much for joining us today. It's been a pleasure to have you on here.
I really appreciate it. For our audience, this concludes another episode of the Society of Critical Care Medicine podcast. If you're listening on your favorite podcast app and you like what you heard, consider rating and leaving a review.
For the Society of Critical Care Medicine podcast, I'm Marilyn Bullock. Thank you for listening.
Announcer: Marilyn N. Bullock, PharmD, BCPS, FCCM, is an Associate Clinical Professor and Director of Strategic Operations at Auburn University Harrison School of Pharmacy. She is also an Adjunct Associate Professor in the Department of Family, Internal and Rural Medicine at the University of Alabama in Tuscaloosa, Alabama, USA, and the University of Alabama Birmingham School of Medicine.
Join or renew your membership with SCCM, the only multi-professional society dedicated exclusively to the advancement of critical care. Contact a customer service representative at 847-827-6888 or visit sccm.org/membership for more information. The SCCM podcast is the copyrighted material of the Society of Critical Care Medicine and all rights are reserved.
Find more episodes at sccm.org/podcast. This podcast is for educational purposes only. The material presented is intended to represent an approach, view, statement, or opinion of the presenter that may be helpful to others.
The views and opinions expressed herein are those of the presenters and do not necessarily reflect the opinions or views of SCCM. SCCM does not recommend or endorse any specific test, physician, product, procedure, opinion, or other information that may be mentioned.