Latest / CRITKNOW / TIGRIS trial

Transcript

TIGRIS trial · CRITKNOW

• Bharath Kumar T V: Hello everyone and welcome back to Crit Now. My name is Bharath Kumar and I'm joined by my co-host Ramesh Venkatraman. Before we get started off, ⁓ reminder that our podcast is available on Apple, on YouTube and on Spotify. If you've been enjoying what we've been doing, please do comment, like, subscribe, share and spread the word. We're very grateful for all of the support and the feedback that we've received so far. Today we're gonna talk about the TIGERUS trial, the trial that evaluated polymix and hemodsorption for patients with septic shock. This is actually a very fascinating trial, but also complex in many ways. So we may end up spending a little bit more time than we normally do in unpacking all of the clinical and methodological nuances that the trial has on offer. With that, I'm gonna pass on to Ramesh to kick us off. Ramesh Venkataraman: you Thanks Bharath, know water trial right kept us busy quite a bit over the last week or so. ⁓ So it's a fascinating trial worthy of the discussion. So and I'm extra happy to do this because ⁓ The senior author of this paper is my mentor who inducted me and enthused me into getting into critical care nephrology. So I owe a lot to him. So my gratitude is to him and ⁓ great work John. So as always and we are happy to discuss this trial here and see where we can go. So I will just give a quick background Bharath and then we go into the literature before. So clearly when we talked about sepsis and septic shock, I think our goal was to try and modulate the inflammatory response. People have traditionally identified several inflammatory molecules and try to address each of them with a specific antidote or antibody if you will and there are at least about 20 molecules that have been evaluated the most famous of which is the activated protein C and none of them seem to have worked in sepsis septic shock. The problem with that is there are multiple pathways that propagate the inflammatory response and if you block one pathway I think the system is smart enough to bypass that pathway and still carry on that it becomes extremely stupid for us to think blocking one molecule or one receptor is going to help. So people moved on to saying can I non-specifically try to address the whole inflammatory response and akin to dialysis can I do something to kind of down regulate the whole inflammatory response if you will and unlike the normal dialysis convection and diffusion does not work for inflammatory molecules purely because of their molecular weight and adsorption ADSORPTION seems to be the primary mechanism by which we can improve this remove these large molecules. And over the decades I think people have started developing various adsorption filters to try and ⁓ various molecules, broadly classify them into selective and non-selective absorption. Selective absorption is what we are going to talk about today, where originally the company from Japan, Toure, started making these polymixing coated fibers that bind to endotoxin very avidly and therefore thought decreasing the levels of endotoxin would attenuate the proximal ⁓ reason for the inflammatory cascade to start and would help patients. There are also several Bharath Kumar T V: Mm-hmm. Ramesh Venkataraman: non-specific adsorptions in the market the. most commonly talked one or cytosorb and then there are newer filters which do multiple things including dialysis like the oxiris. But this study today that you and me are going to talk about Bharath is predominantly using the polymixin endotoxin adsorption. So, just a quick synopsis endotoxin is part of the outer membrane of the cell wall predominantly of the gram negative bacilli and that kind of triggers the inflammatory response that then cascades like a nuclear reaction and goes and bombards all kinds of cells and causes injury. So people figured out that if I can someway decrease the endotoxin levels in the blood by whatever mechanism maybe I can decrease that down the line inflammatory response that is going to hurt the tissues and this was tested ⁓ by using this filter. One of the misconceptions is that endotoxin is just made by Gram-negative bacilli which is not true. A fair amount of Gram-positive bacteria also make endotoxin and therefore it makes a very attractive target for somebody to go and. ⁓ adsorb this molecule and prevent inflammation. That is a quick overview in terms of what we are trying to do, but there has been a fair number of studies done in this. And Bharat, can you just for the audience quickly summarize what the studies so far have shown and where we are at the time of tigress. Bharath Kumar T V: Yeah, absolutely. ⁓ Even how the evidence has evolved is actually fascinating. And we're going to see that, I think, in the discussion. ⁓ one of the earliest trials that evaluated this question, and the journey starts in Italy in 2009. It's called the UFAS trial. This was published in JAMA. The trial essentially enrolled adult patients following emergency abdominal surgery for septic shock. And they randomized patients within six hours of surgery to either usual care or usual care plus two sessions of the polymixing theme adsorption. and their primary outcome and I think that was right, they were one of the earliest trials so they focused on change in mean arterial pressure and a vasopressor index at 72 hours and the secondary outcome was mortality and of course a number of other things. They planned to enroll 120 patients but the trial stopped midway after one of the interim analysis ⁓ and what they saw in the interim analysis essentially is that there was a remarkable improvement in the mean arterial pressure and the vasopressor index but also a substantial improvement in mortality was noted. the trial. versus 53 percent in the usual care arm. Of course, it's debatable whether the trial should have stopped at that point, but nevertheless, I think there were concerns from some of the ethics committees at that point, and the investigators are asked to stop the trial. So a small trial that showed improvement in menatrial pressure and vasopressor score, ⁓ what perhaps was an exaggerated signal towards an improvement in mortality, nevertheless the trial stops. But because it's a small trial, know, obviously obviously we can't treat it as a conclusive body of evidence. the journey then from Italy, most of France ⁓ and the French investigators put together the Abdomix trial, which was from 18 French ICUs published in 2015. ⁓ Again, adult patients following emergency abdominal surgery, but here interestingly the surgeon had to... Once they opened, they had to confirm that there was evidence of peritonitis. And again, the patients are randomized within 12 hours of surgery to either usual care plus polymixin or usual care. alone, polymixing filter, I mean. And the primary outcome was mortality at 28 days here, a larger trial, definitely a more robust trial as well. They aim to enroll around 240 patients and they go on to do that. Their assumptions for the sample size, again, debatable, but nevertheless, at end of the trial, they don't show a statistically significant difference between the two arms. If anything, there's a numerically higher mortality in the polymyx in arm, 27 % versus 19%, but not statistically significant. So that's the Abdomix trial in 2015. The next, yeah. Ramesh Venkataraman: So just to pause you there just for the sake of just clarifying both the UFAS and the ABOMIX study looked at a non-selective kind of a patient group, the only common binding factor being they all had abdominal sepsis. So the UFAS study was predominantly a single center study, I think. And ⁓ the questions that remain in my mind when I read that is, was the appropriate antibiotics given? Was the source control very good or is it just the filter that helps? And the ABO mix was a multi-center study again. And again, the only thing common was abdominal sepsis. But after the negative trial compared to the Bharath Kumar T V: Mm-hmm. Ramesh Venkataraman: previous UFAS, the blood samples of the patients in the ABOBICS trial was they went and looked at the endotoxin levels using a different kind of assay than what we are going to talk about today and they found that the endotoxin levels were not very high. So, one of the assumptions is that maybe ABOBICS was negative because the endotoxin levels were not high enough and hence I think they went on. changing their protocol to the to the next study which is the euphrates. So I just thought we should you take that point how they have evolved in their methodology to fine tune to come to tigris. So over to you Bharath about the euphrates. Bharath Kumar T V: Yeah, yeah. Yeah. Yeah, indeed. ⁓ That's, you know, I think a nice summary of how the evidence evolves. So now the, you know, the journey moves to the United States and North America. So the UFRE test trial published in 2018 and enrolled patients in over 50 ICUs in the US, but also I think from Canada, if I'm not wrong. ⁓ Here, as you said, this is adult patients with septic shock, and they measured the EA level and they included patients if they had an EA level of 0.6 or higher. So this is not just surgical patients. This is all-commerce septic shock as long as they had an EA level with 0.6 or higher. They were randomized to polymix and haem adsorption or interestingly the control arm actually had a sham haem adsorption as well, which of course is very difficult to do. So well done to the investigators. Again, two sessions ⁓ of the polymix and haem adsorption ⁓ of two hours each. And in this trial, they aim to initially enroll 350 odd patients. But after the second interim analysis, they do two things they were advised to increase their sample size and they were asked to enroll a sicker cohort of more than nine so they changed their inclusion criteria to add that and eventually they enrolled 450 patients with 295 in the more than nine group and their primary outcome was 28-day mortality they do not show any difference in the mortality either in the overall group or in the more than nine group the mortality is about 37 percent versus 35 percent overall and 44 or so in the MODs more than 9 group. And an interesting finding in the trial was that the EA levels when they measured it serially did not actually decrease when they followed up serially even after the heme adsorption. So that was an interesting finding. But again, Neuroflitters ⁓ not showing a benefit when they chose this broad group. of patients with an EA level of 0.6 or higher. that, I mean, the same investigators then went on to look at their data in a post-hoc manner because one of the concerns they had was that the assay was not really a linear and we'll talk a little bit about this later as well. The assay is not a linear way of measuring the endotoxin levels. So when the EA's level is more than 0.9, there's actually a much higher increase in endotoxin levels than than what we would think normally. So they were hypothesizing that perhaps polymix in hematosorption would not work at such high levels of endotoxin, more than 0.9. So they did a post-doc analysis where they took patients from the Euphritis trial and restricted their analysis to patients that had a baseline EA level of 0.6 to 0.89. So they wanted to see if it is different in this group. And they had about 194 patients with that EA level. And now when they look at the difference in the mortality, there was about 11 % reduction in mortality for the polymyx synapse, 26 % versus 37%. So in this post-hoc analysis, it appeared to be that there is a benefit for this subgroup of patients that had an EA of 0.6 to 0.89. And this, obviously, they concluded appropriately to say that this is high hypothesis generating and needs to be tested ⁓ in another trial. And that's sort of how I think the evidence has evolved and we reached ⁓ Tigris. Ramesh Venkataraman: So, I think that is a very good timeline of what has happened over a period of time. And I think since you brought the issue about the endotoxin assay, I thought I will take a minute or so to quickly summarize how it is done. So, the endotoxin activity assay EAA, we are not looking at an endotoxin assay directly, there is no real way of measuring endotoxin accurately unless in specific know lab models kind of a thing otherwise it is difficult. So, we look at the endotoxin activity. So, the way it is done is we basically take. the neutrophils from the patient incubate them with an anti endotoxin antibody and depending on the level of endotoxin that is already being present there is an oxidative stress that happens to the neutrophils and they burst and when that happens there is some amount of light that is basically being emitted and this is read by a chemiluminescent assay and then this light is compared to the. theoretically the highest light that you can get with the maximum endotoxin level and then that ratio is what we are basically getting. So, the levels will be anywhere from 0 to 1. And 0 to 0.3 means endotoxin levels are very, very low, 0.3 to 0.59 means they are indeterminate. So, anything above 0.6 you know that there is some amount of endotoxin. So, that is how the 0.6 level cutoff started. The problem here is if the light is too bright, you know that it is 0.9, but you do not know if the endotoxin is just over that limit, it is twice, it is 100 times or 1000 times. So, anything above 0.9 you have no idea to say that this is really going to be helpful with any because if it's sky high no matter what kind of filter you do you will not be able to attenuate that patient. So that's a knowledge that they gained probably retrospectively looking at the euphrates subgroup analysis and when they went to FDA I think the FDA said no. Look, this is a population maybe you can go and look into. Some of the information I'm sharing right now, Bharat is also, you know, in Dr. John Kellum's podcast, which he's recently given into critical care reviews. It's about a 40 minute podcast. He very clearly communicates the whole process of how the study is done. I would encourage our listeners to listen to it, to get insight directly, you know, from the primary investigator. so then they decided, let's just look at the 0.6 to 0.89 group, because that seems to be a reasonable number to do. Bharath Kumar T V: Mm-hmm. Ramesh Venkataraman: and halfway through the Euphrates study, they also found that not only should we have endotoxin level, but you should also have organ dysfunction. If you don't have organ dysfunction and just have the endotoxin level, it doesn't really matter if I just bring the number down without actually affecting anything. So, that's how they refined their science going from a non-selective group of abdominal sepsis to a group which has an intermediate level of endotoxin. Then thinking that endotoxin level with multiple organ dysfunction is probably the right strategy or right patient group to follow. Now they have focused their investigation in tigress to only that subgroup. So this to me is precision medicine that has been beautifully worked out over the last few decades and it's unbelievable lifetime work for a lot of people to get there. It is easy for us to sit and say the study is this and that but then to get here the amount of work that has done is unbelievable. So now we are looking at a specific endo type of patients with a certain endotoxin level with multiple organ dysfunction and that brings us to the TIGRIS trial. Bharath Kumar T V: Absolutely. Yeah. Yeah, I want to sort of ⁓ you made that point very beautifully about how I mean people should know that the EAA actually does it have units, right? So it is not telling you the true endotoxemia or right. It's obviously trying to give you an estimate and the performance is not linear. So it's not that, you know, the relationship is the same at different cutoffs. So I think that that is important to, I guess, emphasize. And I agree with you. This is, you know, how beautifully science also evolves, right? sort of trying to improve our understanding step by step but obviously it takes a lot of time a lot of commitment and dedication so a great work by everybody that's gone on that journey so Ramesh Venkataraman: Amazing, yeah. So now that brings us to the tigris study Bharat. So I'll just quickly tell what was done in tigris study. Now that they needed to look at the specific endotype, they said that we will recruit patients who exactly meet this euphritis subgroup that showed benefit. they basically said we will enroll patients with septic shock, shock defined as at least 0.05 or 0.1 mikes per kilo per minute of norepinephrine for at least two hours, but not more than 30 hours. Because if it's too late, then probably a therapy is not going to work. So they wanted to catch in that window. You should have multi organ dysphonia. dysfunction as measured by a multi organ score of more than 9 and then you should have the endotoxin level as we spoke about between 0.6 to 0.89 and then you should have one of the organ dysfunctions including ventilation, thrombocytopenia or one more thing which I am blanking out on right now. So they had very specific criteria to select patients. The interesting thing is they screened almost 15,000 patients to enroll about 150 patients. So and they randomized patients as two is to one ratio, which we will talk about the pluses and minuses of that. So two patients go into the intervention arm and one patient goes into the standard group arm. Patients were randomized to two treatments of the polymax in hemo adsorption, two treatments of two hours each, at least about 24 hours apart. So the average time is about 22 hours or 24 hours apart along with standard of care. Standard of care is whatever the sepsis guidelines would say like fluid resuscitation, appropriate antibiotic source control, all of that. The comparison, the control group only had the standard of care without having the polymaxin filter. So, you may wonder why two treatments, why 22 hours apart? There is some rational, although it is not a very tight rational that has been proven beyond doubt. The rational is not Normally you would do these heme adsorptions at about 100 ml per minute flow. Unlike dialysis you want ⁓ adsorption, you want to adsorb the surface so the slower the blood flows the adsorption gets better. So the volume of blood flow would be about 100 ml per minute. If you do 100 ml per minute for two hours on an average you would do four to five blood volumes of adsorption on a filter that should be able to adsorb a good proportion of the endotoxin that is circulating in the blood. Bharath Kumar T V: Hmm. Ramesh Venkataraman: the second treatment 24 hours later, that can be again reaccumulation of endotoxin that is released from gut and elsewhere and that can again come back to the blood. So you want to go and capture that surge happening roughly about 24 hours later. And if you look at it, all the studies have done two treatments 24 hours apart using this as the primary rationale. So this was what was looked at and their primary mortality was 20 was primary outcome was 28 day mortality and that was the primary outcome. They had a lot of secondary outcomes and one interesting thing and I think this is where I will need your help Bharath is that they didn't do the usual frequentist approach of analysis. They have done something called the Bayesian approach of analysis in terms of trying to determine whether their outcome was positive or negative. So I will stop here Bharath and I think you are better qualified than me in trying to know elaborate on frequentist and versus Bayesian method and why one would use one bus. as the other. Bharath Kumar T V: Yeah, thank you. I also want to sort of point out, as you said, that this is an enriched population that they are including into the trial very specifically and trying to be as close to euphoritus, of course, as possible. And there are reasons why they did that. So coming to the Bayesian versus frequent test, I want to make maybe two points. Obviously, you know, a detailed explanation will be beyond the scope of this podcast. But very quickly, the first one is the nature of the hypothesis testing. So we all are familiar with the null hypothesis and the alternate hypothesis. The null hypothesis says that there is no difference between two treatments. And the alternate hypothesis says that there is a difference, either better or worse or whichever direction. Now in a traditional frequent test analysis, what we doing is that we are estimating the probability of seeing the results that we eventually get or something more extreme than that. given if you assume that the null hypothesis is true. So this is a bit of a convoluted way of looking at the data, right? So what is the probability of getting results that I got or something more extreme where the null hypothesis to be true? And if that probability is less than 5 % conventionally, we reject the null hypothesis, right? So that is what we normally do. But what you and I care about is what is the probability that my hypothesis is true, right? That is what we are looking at. We want to look in the other way, not this convoluted way of going about things. And Bayesian approach is actually asking that question. What is the probability that polymix and haem adsorption saves lives? Right? That or my hypothesis is true. So that is the fundamental difference between a Bayesian approach and a... Ramesh Venkataraman: to pause you and frequent is our probability tells you the probability of rejecting the null hypothesis here the probability is probability of the null hypothesis being true the original hypothesis being true. Bharath Kumar T V: Correct. But also the way we look at it, right? We look at the data and say, what is the probability of getting such an extreme data or that kind of data if the null hypothesis were to be true? And when it reaches a certain threshold, we are saying, OK, we can now reject the null hypothesis. But here you are sort of coming at the question more in a prospective or in the correct direction, where you're saying, I have a hypothesis. What is the probability that my hypothesis is true? Ramesh Venkataraman: Hypothesis. Bharath Kumar T V: So that's a bit of a flip in how they look at it. But I would argue that as clinicians, that is what we are obviously interested in. That is also how we think. Now the second important point when we talk about Bayesian analysis is something called the prior. And this is a concept that we need to understand if you have to understand Bayesian methods. So what we do in a prior is we don't look at the data that the trial generates in a vacuum. We say, OK, what is the prior body of evidence on this? Now I have done my trial and there is some data. Now how does that trial data update the prior probability to give us a new probability? So that is what Bayesian approach is doing. And again, it Ramesh Venkataraman: So it's very fair, mean as clinicians that's what we do when we order a test, right? We look at what is the pretest probability of a test, then we get the test result and then integrate it with the pretest probability to say what would be my post-test probability of a particular diagnosis, right? This is clearly what we do with PE for example, pulmonary embolism, right? So I think that people would be very easily relatable to and I think this is the same way that Bayesian method works if I understand it right. Bharath Kumar T V: Correct. Yeah, exactly. Absolutely, that's the perfect example, when we're a test, as you rightly said. So as clinicians, we actually are inherently Bayesian. It is only when we were doing the analysis we were thinking of these frequent test methods. So you have a prior, and now the prior can be of different types. So if you have a lot of data before a particular study, can use that information as your prior, and you call that an informative prior. Now you may also start sometimes on a blank state where you basically say you have no prior, no prior knowledge of a particular issue which is an uninformative prior or there is some data but not great quality where you will call it a weak prior. So these are some of the terms we will come across. Now in the case of tigress what they did is obviously there is prior data but more importantly there is prior data from a trial which had endotoxin measurement. So they rightly use that data to inform their prior, the euphrates post hoc analysis specifically. So this is an informative prior that they do. Then they run their trial and then they update the results based on the posterior probability. Now what they do slightly differently though is they include the euphrates data into their analysis itself, which is a bit different to what many of the other Bayesian trials have done. So in some ways they are pooling the data from two trials. So what a meta-analysis would normally do, but here they're doing it in the context of a trial itself. So they're taking some of the euphrates information and then they're adding it to the TIGRAS information and then they're giving you a posterior probability. So that's, it's an embedded approach. So the prior is embedded in terms of including the data as and they are able Ramesh Venkataraman: So just to to just to interrupt you the Bharat, I think the rationale for that is the costs involved in conducting a much, larger trial. You want to enhance the power. We are very clear that we want to look at a specific endo type of population, which we have already shown that there is a benefit. So enroll similar patients, pool the data so that the overall power would go up without additional costs and logistics that that the trial investigators have to go through. That is a valid reason considering a very very niche area with a expensive therapy like mean relatively expensive therapy like hematopsy. So I think that was the rationale for that. No sorry over to you. Go ahead. Bharath Kumar T V: No, no, that's absolutely right. Yeah, it is a difficult trial to do. So obviously you want to take data from prior trials if they are in a similar population. And I was going to come to the point to say that they were able to do that because they thought that if they enroll a similar set of patients, the information is sort of exchangeable. So you can actually pull that information. And we'll again maybe talk a little bit about that. They obviously did not take all of the information from Euphritis. They attribute a certain weight to it. And we'll talk a little bit about this later. is the weight that they give. And then you know that tigress has done, you interpret the tigress results in the context of what has previously been done. Now a couple of other points for a sample size, they choose, they assume the control group event rate or mortality for about 48%. I'm not sure where this came from because that was not the post hoc euthanitis data. So that was not clear to me. ⁓ But Overall, they do a number of other sensitivity analysis that try to check their model assumptions to be confident that the results they get are robust. And we have spoken about sensitivity analysis in one of our prior episodes. So from an analytical standpoint, these are the major points that I wanted to bring out. A Bayesian approach, I want to talk about the priors and how they incorporate the prior information. And of course, they do a number of sensitivity analysis to test the robustness of their findings. And when we come to the results, we will see that it sort of did play out. the results were robust to many of those ⁓ approaches. So maybe we can get into the results. Ramesh Venkataraman: Yeah, so I'll just look at the results when I look at results. just look at. these patients similar to what I would. Yes, assume that I have an endotoxin assay and I am able to get the patient's assay level, but clinically, what is the demographics? Most people are about 60-62, which is the population that I would see. Roughly 50-50, 45 and 45 are the male-female ratio, is fine. The thing that struck me the most is the Apache 2 score. Apache's 2 score was 32. 32 is extremely high. For those of you who are not familiar, familiar by just looking at a number and figuring out the severity, 32 is very high. To basically quote the Zyggr study, the Apache 2 in the population where activated protein C was evaluated at an Apache of 25, they felt that that is the subgroup that may benefit more because of the severity of illness. Here we are talking about 32. So, extremely sick patients who had multi organ dysfunction, so clearly their mortality was very high. like we do in clinical practice one third of them had Bacteremia that was documented the rest were culture negative and among the Bacteremia there was both gram positive and gram negative as we would expect because endotoxin is made by gram positives too. Almost 90 percent and more had an R-Epinephrine dose of more than 0.1 mikes per kilo per minute again significant amounts of shock that we had to have for enrollment. So with that kind of a thing most people got the therapy they had to get roughly about 100 patients in their intervention arm and 50 patients in the control arm considering it is 2 is to 1 and the 28-day mortality if you look at the probability by Bayesian method they say there was a 95 % probability that heme adsorption using polymaxone is going to improve survival. But if you look at a frequentist approach and just look at their numbers, the mortality was 38 % versus 45%. 28-day all-cause mortality between the two groups, there was an absolute risk reduction of about 7%. So that was there. That was not enough for the investigators and obviously for the FDA and then they went on to 90-day mortality evaluation also. The interesting fact that happens here Bharat is something counterintuitive. Normally you would say, no, maybe I'm doing something, no, to get them past the 28-day mark and most of them die after that but actually the opposite happened. The robustness of the trial results was even better when they looked at 90-day mortality. The probability went up to as high as 99 % that this therapy is going to help and the absolute risk reduction was almost 17 % if I have my numbers right 43 % in the treatment arm and 60 % in the intervention arm which you would expect 60 % in the standard arm and 43 % in the intervention arm considering that the Apache was very very high. So the 90 day signal was even more profound compared to 30 day signal and that again makes it more robust and like you said they had done a lot of modeling and sensitivity analysis and the robustness in the 90-day mortality never changed. It was consistently robust despite however you looked at the data and whatever analysis you did and that makes me feel good that this is not a therapy that is just tiding over the crisis but has some impact at reasonably long-term mortality. The interesting fact is after getting all this, when they went to FDA, FDA said just get us more data. So now the investigators are actually looking at even longer term outcome, maybe a year or six months. mean, I think it's a year if I go by what I've read out in the web. So they're looking at one year outcomes to see if that would be strong enough to convince FDA to approve this device for this particular purpose. So that's the quick overall view of that study Bharat. So what are your initial thoughts after I've shared methodology with you. Bharath Kumar T V: Yeah, I want to just maybe one comment and maybe a question to you. One was I think they very nicely presented in table one the data of both tigress and the cohort that they chose from Euphritis to give us a sense of are these patients similar? And broadly, they were similar, I think, right? That is important for us to be, yeah. Ramesh Venkataraman: Absolutely, they were similar. I think you brought a valid point. I should have said that in my discussion, but they're pretty much matched across all the things they've looked at age six, Apache, the MOT score, the SOFA score, the percentage of bactremia, cumulative vasopressor index and endotoxicity assay. So almost across all of those domains, I think the euphritis and the tigress, the patient that they took from euphritis and the tigris patients actually quite well match. So, I think that is very commendable. mean they are pulling in data is reasonably valid considering that they picked very similar patients. Bharath Kumar T V: Yeah and ⁓ again more as a question they also did check you know different weights to the euphritis cohort right yeah they did yeah. Ramesh Venkataraman: So I'll just, ⁓ mean from my understanding Bharat, like you said, they have to take patients from euphritis to enhance the power but euphritis was published at least almost 10 years I mean 8 to 10 years ago compared to tigress. So even though you have matched everything possible endotoxin assay, the age, the MOTS score, the vasopressor index everything there is always going to be some amount of change in the way sepsis is managed and that itself would have had an effect on mortality and the way we practice now. So it is not a simple addition of patient one as the first group patient one versus the group in the second group and just say we will clump and do it. I think this is where you have to use a little bit of common sense and what the investigators have done is we understand this limitation. that we are pooling similar, even though they are similar patients, we are pooling it from a study in the past. We don't know if they, if we can take it in the face value. Therefore, we are only going to give that information 75 % weightage. Why 75 % is very hard. There is no previous president for anybody to say this should be the weightage or this should not. I think the investigators took a very pragmatic approach to say, I will do 75 % weightage of the euphritis data. And when you talked about sensitivity analysis they have changed this weightage say what if your frit is only 50 % what if you produce was 100 % and they have done lot of robust analysis and and the 90-day mortality advantage still remains despite the weightage, ⁓ the relative difference in the weightage that they are given over a period of time. So that's what the weightage from the euphritis means when you read the paper. It can be a bit confusing, you what is a prior, what is a weightage, you very nicely talked about the prior, but this is the rationale for weightage. And again, very clearly explained by John in his podcast too. So I think listening to this and that would be a very comprehensive way of understanding this trial. Bharath Kumar T V: Absolutely. As you rightly said, think there isn't any clear answer to say how much we should penalize the prior data. So they went with 75%, but they also give you results for a variety of assumptions. And again, as you pointed out, the 28-day mortality is not as robust to those variations, but the 90-day mortality seems to be, the outcomes seem to be pretty robust to those variations. Yeah. Ramesh Venkataraman: So it's really interesting, like most, if you look at the, like a lot of therapies which have looked at the initial phase of sepsis, I remember the first Emmanuel Rivers early goal directed therapy study, you're resuscitating somebody for six hours, you're showing a difference in 28 day mortality, but the survival advantage remained at 60 days. No, so whatever you're doing early seems to have a long effect and, that, and And the most common explanation thrown out by experts is that, that whatever intervention you're doing to attenuate the inflammatory response or optimize your hemodynamic response is going to avoid more and more organ failure later on. And that translates into better outcomes later on. Obviously, a shock reversal may be better early on, but I think the organ rates are, organ failure rates are much better in the treatment at a later phase because of what we did early. And interestingly, even here if you look at the 28 days, most patients in the standard group still remain in the ICU at 28 days even if they survived versus a good proportion of patients where out of the ICU in the treatment arm. So, that again indirectly tells us maybe you are coming off organ support much more faster with an earlier intervention like the polymaxine de adsorption, but it is encouraging to see. more robust outcomes as you go longer because then you definitely modulating the inflammatory response and maybe favorably tilting the balance towards better organ preservation. Bharath Kumar T V: Yeah, absolutely. So I want to talk a little bit about what I thought were big strengths and then maybe some concerns from my perspective. I think from the strengths, ⁓ as far as I know, this is the first trial in critical care to formally test a signal seen in a prior subgroup and do it in a robust way, right? In a very sort of a precision-enriched approach. very difficult trial to do as we have seen you know between euphoritis and tigress they screen over 35,000 patients to include about 350 patients so very very hard to do this kind of a trial so I think ⁓ terrific work And as we said before, the whole series of studies that have been done and how it has evolved. My concerns, I had a few. ⁓ One was, yes, table one shows that the populations are similar. But obviously, these patients have been enrolled in very different eras in terms of how critical care has evolved. So where the patients truly exchangeable is obviously up for discussion. They mitigate that to some extent by doing all of those sensitivity analysis, but one can never be 100 % sure that these are really the same set of patients that can be pulled together. So that was ⁓ one concern. The second sort of a concern I had was that when I went through a very long supplementary appendix, it didn't appear like day 90 mortality was initially pre-specified. It looks like it's something that came up later. I'm not sure at what point it was added. I gathered that the FDA asked them to add that and also look at longer-term outcomes, but it's not clear entirely when it was added. And obviously, we see that the longer-term outcomes is where the results are more robust. It would have been good to get some clarity on that. The third concern that I wanted to point out is the 2 is to 1 ratio. Now, normally, we use a 2 is to 1 ratio when we want to, when we have a reasonable certainty that something is working, but we want to obviously confirm it, and then we... want to ensure that a larger number of patients get what appears to be a beneficial treatment. Now, starting the polymyxin trial, I'm not sure that we were entirely clear that polymyxin was, the hemodsorption was beneficial. So I'm not sure what the rationale for going with a 2 is to 1 ratio was. And then when you do the 2 is to 1 ratio, it's generally a less efficient ratio. It reduces your statistical efficiency. And if you have a lesser number of control arm event rates, it reduces the precision. so there are a lot of, if the trial is a small trial or it stops early, it biases the estimates. So there are some problems with that two-to-one approach. I recognize why they took that approach, but I still think that it brings a little bit of problems to the, yeah. Ramesh Venkataraman: No, I think it was just a balance between being logistically pragmatic about what could be done and pure science, right? Sometimes, you may have to do what you think is doable. mean, at least we got these kinds of results now, maybe one is to one, I don't know, it would have taken even more longer. Now they had to screen 14, 15,000 patients to get 150 over so many years. And that is going to be a big. problem for the investigators. think they took a very pragmatic approach and it very understandable. You have been working on a technology for two decades and you still trying to get it into the bedside for routine use and I think that you just want to be able to get the results out and be of some use and I think this study has definitely told us how research should be done in a very heterogeneous way. kind of a population like septic shock, right? You slowly define your subgroups that you're going to aim your therapy at, use that subgroup to try and get an answer. And then yes, in en route, like you said, the two is to one, all of those limitations do happen, but I still think it is kudos for the investigators to have been very perseverant for so many years to get here to where they got there. Yes, we also need to be cognizant of the limitations when we interpret the results. Bharath Kumar T V: Yeah. Yeah. Ramesh Venkataraman: But if you ask me what would I do tomorrow, I would say if I get a patient where I am able to do an endotoxin assay at the bedside, which we used to for a short while in India, I am sure that it will start penetrating soon again here. If I have the same EAA assay and I have a patient with EAA level of 0.6 to 0.9, who 0.89 and who has MADS, multi-system organ failure, who fits this criteria, I would certainly offer this therapy to that patient. I can't ignore an absolute risk reduction of 99 % probability of improved survival in nine months, in 90 days, and an absolute risk reduction of close to 10%, right? And Euphrates also, and so that becomes a valid point. And I think the rest of the stuff, yes, I am skeptical a little bit about industry funding the trial, even though industry had nothing to do with the analysis and writing of the manuscript. The numbers of this trial are very, very small. And there is a non significant trend if you look at it from a frequentist point of view. So, all of those makes me a little septic. On the other hand, to counter all this I also feel that the 90 day mortality even though it was not the primary outcome consistently showing benefit irrespective of the type of sensitivity analysis and weightage they have given makes me believe that this therapy would more likely be helpful than harmful. for the patient who fits this very, very narrow criteria. So I think that is something important. Bharath Kumar T V: Now, rock it. Yeah. No, mean, my take home is, I would say, fairly similar to say that if you meet all of those criteria, I would certainly offer, you know, but explain, obviously, the potential benefits, possible risks and the cost implications, especially in our setting. But I would certainly offer in that group. ⁓ My one concern, obviously, with all of these interventions is the indication creep, right? Because when it gets applied, people necessarily don't follow what the trial did. But that's true for many other therapies as well. Ramesh Venkataraman: Pause. Bharath Kumar T V: ⁓ And we need to be very wary, right? Who do we offer to? And this is not every septic shock. This is not everybody who can measure an EA on, right? ⁓ This is this very specific group. And the other, I mean, maybe another uncertainty which future research needs to answer that occurred to me is that we still don't know what is the optimal dose, right? How many sessions ⁓ and, you know, is one year level measurement initially enough? How do we sort of, if the patient is not improving, do you repeat and recheck? So there are those uncertainties that remain. But having said that, maybe I also want to say that it's unlikely that there is going to be a lot of additional research happening in this space given how challenging it is to know enroll patients. So I'm also obviously cognizant of that. ⁓ So those were sort of my big thoughts and takeaways. Was there anything else that you thought we should Ramesh Venkataraman: No, no, I think that all the points you make are very valid. We've got to be extremely cautious in terms of who we are going to give it to and whether two treatments is enough versus more. And if you look at studies that have evaluated the endotoxin activity assay itself, most of those studies have looked at at least two levels, not a single level. And as you know, endotoxin activity assay depends on neutrophil function. If you have neutrophil dysfunction, immunoparalysis, somebody's already had partial treatment, for sepsis all of those can decrease the indirect endotectivity activity as a like I said the neutrophil burst can go down and we can get a falsely lower result. So that's something that that we need to keep in mind. And the other point I thought that's worth sharing Bharath is that in a trial to recruit somebody into a therapy within 30 hours even though it sounds like a long period is extremely difficult right you need to come in get them resuscitated they need to meet certain criteria get into the mod score get a EAA and then randomize get the therapy started within 30 hours is unbelievable. And I think that would have restricted the investigators in terms of, I mean, I'm sure a lot of exclusions happen just because of that. And John does mention that in his podcast too. But in real life, I think we may be able to get a little bit more patients than the one person. So I think one person sounds too pessimistic, right? Of all the patients screened, only one person got therapy. So the take home message is not that one person of the septic patients is going to fit criteria, but patients who fit the criteria is going to be extremely narrow, but probably much more than what the study showed because study conditions are always lot more rigorous and very strict compared to what we would do at the bedside. that's an additional thought I think we need to bring out to the audience so that ⁓ I think they think through all the angles before they actually go ahead and say yes or nay for this particular therapy. Bharath Kumar T V: Yeah. No, that's great points again. And I guess our broad take-home is if the patient meets the inclusion criteria in the trial, it's worth a shot, obviously, after counseling the patient and the family. But overall, again, I agree that it's a great trial, very tough to do. Congratulations to the investigators. And on that note, we will stop here. Thank you so much once again for tuning in, for listening in. And as I mentioned initially, please do comment, subscribe, like, share, and spread the word. Thank you so much. Thank Ramesh Venkataraman: Thank you all. Bye bye.