Transcript

Andromeda-Shock-2 · CRITKNOW

Bharath Kumar T V: Hello everyone and welcome to the first episode of CritNO. My name is Bharath Kumar. I'm an ICU doctor and researcher based in Chennai. My research interests include clinical trials and research methodology. Our podcast will focus on a variety of themes and papers related to the practice of evidence-based critical care. To start with, we will try to air one episode every fortnight. Joining me on this journey is my co-host, Dr. Ramesh Venkatraman. Ramesh Venkataraman: Thank you Bharath, this is exciting. think this is a good initiative and hopefully we will have fun doing it too. So in this podcast today, we thought we'll take a recent study, talk about some methodological nuances and then the clinical relevance. And I thought the most recent study that has made a lot of noise is the Andromeda 2 shock trial. So I thought we will focus on that today Bharath. Bharath Kumar T V: Absolutely. Yeah. So we'll try and keep the tone of this podcast conversational as much as we can. So we'll get started. So Dr. Ramesh, perhaps you can start by talking about ⁓ this concept of goal directed resuscitation and how it has evolved ⁓ over the years, leading, of course, to the rationale for this particular trial itself. Ramesh Venkataraman: Yeah, I think that will be a great start Bharath. So ⁓ till about 2001, I think when a patient showed up to the emergency room with sepsis, I think the resuscitation happened ⁓ in no specific order or with no specific structure to it and people used to do whatever they thought was best for the patient. The first study that looked at A specific target for resuscitation was the early goal directed therapy resuscitation by Dr. Emmanuel Rivers from Detroit in 2001. It was a single center study where he had specific goals of the patient. For example, a target mean arterial pressure, a target urine output and the primary target of resuscitation in that study was a central venous oxygen saturation SCVO2 using a continuous catheter. And he showed that if you are able to target specific physiological variables over the first six hours and optimize the SCVO2 to keep it above 70%, the mortality was very good. was not only immediate mortality, but also delayed mortality was very good among these patients. So that created a lot of buzz around that time. And I think people started following some of the ⁓ methodology that was used in the study in clinical practice. That study although raised a lot of awareness about how non-structured a resuscitation was, also was criticized for a few things. Number one, the primary target of resuscitation was central venous oxygen saturation. As we all know in sepsis, one of the big physiological derangements that happens is impact tissue extraction and utilization. So you would expect the SCV voltage to be high rather than low in most patients who showed up in sepsis. And that's what rest of the world showed. So if they see what is already about 70 doesn't mean somebody's already resuscitated and you can pack your bags and go home. So that was a very ⁓ important message that came out. Second, there was a lot of things done in the study like blood transfusions, adding nitroglycerin for microcirculation, where people felt a little uncomfortable in terms of the variables that were used. So. And Detroit is also unique because of the demographics of the patient population there. And most of them probably do not have ⁓ health insurance. And they probably present to the emergency room late. So you can't extrapolate that care model to the rest of the world. So I think there are three parallel studies done. But it almost took a decade of all these discussions to happen before the next three major studies that were done in three continents. One in US, the process study. Bharath Kumar T V: Mm. Ramesh Venkataraman: One in Australia the array study and one in UK the promise study all three of them used various forms of goal directed resuscitation to see if they would Show improved outcomes in patients requiring sepsis resuscitation all three of them were negative trials And then there have been a pooled meta analysis that has also subsequently showed lack of benefit with early goal directed therapy So obviously all of us here wouldn't believe that ⁓ early goal directed therapy doesn't help would be a very wrong conclusion for all of us to have because we need to have specific goals to resuscitate and we need to do it early. So there is no questions about the logic of early goal directed therapy. The question that then that arises is one specific goal or a fixed protocol appropriate for every patient who walks into the emergency room. So maybe that's the reason why these studies were negative. And the second thing is between the first study, the Emanuel Bharath Kumar T V: Mmm. Ramesh Venkataraman: reverse first single center study and the subsequent multi-center studies a decade had gone by during which probably we got better even during regular resuscitation that the separation between the control arm and the protocolized arm was probably not significant enough to capture mortality benefit. So the first question is should my resuscitation be based on targets but the target should be individualized based on physiology. So that is one question that we need to address. Second, if SEVO2 was not a good target, then what else can be a good target? SEVO2 is an indirect marker of global oxygen delivery. And that's what the argument was. But clearly, we talked about the short fallings of SEVO2. Then the question is, can I use lactate as a marker? The problem with lactate is it's not widely available everywhere immediately. And two, Lactate can be confounded by other things, most importantly the sympathetic drive that happens during any kind of illness where lactate is a marker of sympathetic drive and the severity of the patient's illness rather than actually hyperperfusion alone. In that context, ⁓ a few years ago, the Andromeda group did a earlier trial, the Andromeda 1 trial where they compared a simple backside clinical tool, capillary refill tying. They used 3 seconds as the cut off and they compared with lactate, serial lactate measurements as a target of resuscitation. Without going into too many details of that study, they broadly showed that a capillary refrid time guided resuscitation was as good as lactate guided resuscitation in all patients with sepsis. So, now we have two different pieces of puzzle. One, do I need to individualize my resuscitation for septic patients and two, can I use a tool that is widely available that is a marker of global perfusion and that can be used in any setup irrespective of the resources available which is capillary refill time that has been validated in the previous study. So, the Andromeda 2 study has basically used these two rational and basically said can I do personalized resuscitation and can I use CRT as a target and can I basically evaluate and see whether if I do that kind of an approach I can improve outcome. So, broadly That's how I see the literature emerging over a period of time that has got us down to the Andromeda 2 shock trial. Bharath Kumar T V: Yeah. Yeah, no, that's a very ⁓ scholarly, brief, and comprehensive summary of, I guess, how the field has evolved. So thank you so much. I also want to quickly sort of make the point that to listeners, it again underscores some of the caveats of adopting strategies or interventions based on a single center ⁓ study. So subsequent trials obviously did not bear out those early results. And of course, you've discussed a variety of reasons for why that might be the case. So coming to the trial itself, Andromeda shock too, ⁓ maybe I'll talk very quickly about, you know, what they've done. ⁓ Essentially the research question that they were asking is, ⁓ is a personalized hemodynamic resuscitation protocol or strategy superior to usual care? ⁓ And usual care here refers to centers using their local or institutional protocols or international guidelines to direct care. ⁓ This trial enrolled adult patients ⁓ over the age of 18 ⁓ who were presenting with septic shock and the way they define septic shock is that ⁓ you had to have suspected or confirmed infection with a lactate more than two and ⁓ requiring vasopressors to support a map more than 65 after having received at least one liter of fluids. So I would say a pretty standard definition for entry. The trial was conducted at over 80 centers ⁓ predominantly in the Americas and Europe, but also a few centers ⁓ in Asia. ⁓ The comparator arm, I think we said is the usual care which we've discussed. So I want to talk a little bit about the intervention itself. ⁓ So in the intervention, ⁓ again, those that have read the paper would be familiar with what they did. Essentially, they had two tiers, tier one strategies and tier two strategies, both, of course, ⁓ using CRT as the ⁓ target endpoint. ⁓ Essentially, after this one liter of fluid, if you get into the trial, they first check your CRT. If the CRT is already normalized, you would go to periodic monitoring. If the CRT had not normalized, they would check the pulse pressure. If the pulse pressure is less than 40, you would get a fluid, you know, they'll check for fluid responsiveness, do a fluid challenge. If they were responsive, they would get fluids, and then they'll again check the CRT. If the CRT has normalized, you go to periodic monitoring. Now, if the pulse pressure was more than 40 to start with, they would check the diastolic blood pressure. If that was less than 50, they would adjust the norepinephrine to target a diastolic blood pressure more than 50 and then recheck the CRT. So, if this normalizes the CRT, you go to periodic monitoring. If this does not normalize the CRT, you sort of come back and check for fluid responsiveness. Dr. Ramesh, does tier 1 make sense? Ramesh Venkataraman: TR1 kind of broadly makes sense. The points that are reinforcing is pulse pressure is a surrogate for stroke volume. That is the physiological assumption that most of us make. So, if the stroke volume is low, the first assumption is we want to make sure that they are appropriately filled. So, that is the reason why the pulse pressure is low, they check for fluid responsiveness and then go ahead and give the fluid. But once they are not fluid responsive, then the next step obviously would be to correct any vasodilation that happens and then they use diastolic blood pressure again. I think the rationale is to keep it as low key so that people can use it all around the place. then, Bharath, can you elaborate more on tier two and the rationale for doing the two hemodynamic challenges? Let's say they have gone through the pulse pressure, they've gone through the diastolic blood pressure and still your CRT is low. Bharath Kumar T V: Mm. Ramesh Venkataraman: Then they specifically in tier two deployed two challenges, one with a higher map and other with an high note drop. Could you just elaborate on the rationale for those two and how it's done, please? Bharath Kumar T V: Yeah, sure. So in tier two, they actually start off with an echocardiographic assessment. ⁓ if, of course, if there is evidence of cardiac dysfunction, then you would target your therapies towards that based on LV or RV dysfunction. ⁓ But if there is no echocardiographic evidence of dysfunction, you again go back to fluid responsiveness. If you're fluid responsive, you give fluid. If you're not fluid responsive, ⁓ the next step they check is, you previously hypertensive or not? Ramesh Venkataraman: you Bharath Kumar T V: This of course sort of draws on from the sepsis spam study, right? Where at least in the hypertensive group, targeting a higher mean arterial pressure seem to reduce the need for renal replacement therapy, if not any other outcomes. So here they do what is called a MAP test where they target a mean arterial pressure. more than 85 for one hour and then they check the CRT. If that normalizes the CRT, they continue the same strategy for six hours. If it does not normalize the CRT, then they try a dobutamine ⁓ challenge or a dobutamine test. So here they would give dobutamine at five mikes per kg per minute for an hour and again, recheck the CRT. If the CRT has normalized, you go to periodic monitoring. If the CRT has not normalized by this stage after all of these therapies, you sort of move to rescue therapies such as steroids. extra corporal adsorption and so on and so forth. Again, I want to quickly check, ⁓ Dustier 2 makes sense to you. What are your thoughts? Ramesh Venkataraman: Yeah, I mean, I think they did a couple of things. Clinically, they both make sense. Obviously, somebody where you have adequately volume resuscitated, improved your mean atrial pressure to the traditional standards of 65 millimeters mercury and the STRP is still low. They have held on to the subgroup of sepsis spam patients who are hypertensive and they're trying to raise the map. The last thing you could do is to improve contractility, especially if they have bad LV function. They've done an echocardiographic analysis and I think all of us would do the same thing at the bedside. What I liked about both these challenges is that they kept it to a brief minimum of about one hour because both these interventions are not without harm. think that's something that I want to bring it to the audience here that both raising MAP up to 85 has its own adverse events and then arbitrarily giving an inotropic dobutamine would also probably not be a good idea if there is no response. So I like the fact that they tried it only for one hour and then if it worked then it makes physiological sense to continue. If it didn't work, they moved on to some other strategy and aborted these two trials. So from a physiological and a pragmatic point of view, I think they have done whatever I would do at the bedside at this point of time. And I like the fact that they kept it to one hour to minimize harm. Bharath Kumar T V: Yeah, absolutely. I think it has that element of sort of safety built into it even as they as they try to, you know, try these tests. So I think that's fantastic approach. Dr. Ramesh, that sort of brings us to the outcomes that they chose. ⁓ And obviously in this trial, they're using something that we don't always encounter, but it's becoming common, I guess now, which is this idea of composite outcomes. ⁓ So should we talk a little bit about composite outcomes and what they did? Yeah. Ramesh Venkataraman: Yeah, I think it would be a good prelude to address other nuances in the study. So traditionally when we look at any intervention, we just look at mortality, a 28 day mortality, 90 day mortality. Usually most studies look at that because it's straightforward, easy and quote unquote the gold standard to say if somebody improves the outcome. But the problem is that there may not be too many mortalities and to find a difference between mortalities, you need huge number of patients and to do a trial, like this with tier 1, tier 2 and multiple interventions would just be very, very difficult across several centers. So the only way to get the maximum yield is to club two or three outcomes into one and then traditionally use it. And most people have used a composite outcome. The commonest example that comes to my mind is they make 90 at 90 days, whether you are dead, whether you have AKA with a creatinine rise or whether you are on renal replacement therapy, all three are clubbed together so that the number of events is so high that the power of the study can be achieved with less number of patients and less costs and complexity involved. The problem with composite outcomes is that say for example the same make 90 I am clubbing mortality with a bump or a persistent elevation in creatinine obviously I would more than happy to live with a persistently elevated creatinine than being dead. combining those two outcomes clinically doesn't make sense. Bharath Kumar T V: Mm. Ramesh Venkataraman: And number 2, sometimes one may go in one direction, the other may go in the other direction and that kind of confuses the analysis. So, what do I do now kind of an issue happens. So, what the next level people have tried is to do something called a hierarchical outcome. will say okay, death is the most important outcome followed by renal replacement therapy and then followed by creatinine. That kind of hierarchical outcomes make sense because the weightage given to each of those outcomes would be different. So, that is how. composite outcomes have evolved over a period of time, but they have gone one step further and used a new technique called wind ratio. And Bharat, do you want to take the issue further and just discuss what wind ratio is and how we should look at it? Bharath Kumar T V: Mm-hmm. Yeah, absolutely. So ⁓ that's, again, an excellent summary of composite outcomes. And of course, we briefly touched upon the pitfalls of composite outcomes as well. ⁓ So when you're analyzing composite outcomes, there are different statistical approaches. And one of the ways to do this is to use the bin ratio, especially when you have hierarchy inbuilt into the composite outcome. listeners again are going to be familiar with odds ratios and risk ratios I think. So win ratio is essentially another way of evaluating if an intervention is effective or not, right? ⁓ And this is specifically in the context of a hierarchical composite outcome. So what we do with win ratios is we take every participant in the intervention arm and we compare them with every participant in the control arm and then we do this in a hierarchical manner for each of the outcomes. So first we compare for depth then for the duration of organ support and then for the length of stay. Now, when we come to the first outcome of death, we compare the number of wins, losses and ties. So what I mean by that is they will check, for example, for death, they're going to check if the intervention arm patient survived and the control arm patient died. So if the intervention arm patient has survived and the control arm patient has died, that's a win for the intervention arm. Obviously, it can be vice versa, in which case it's a win for the control arm, right? They can also be ties. So if both patients die, that's classified as an early tie and they sort of are not contributing to the to wind ratio analysis beyond that. And if both of them are alive, now we have a tie that has to move to the next stage. So at the next stage, now the participants are compared for the outcome of duration of organ support. And obviously, if you take it as organ support three days, the higher the number of days you have a win. Ramesh Venkataraman: Thanks. Bharath Kumar T V: the lower the number of days you have a loss. So they again compare this for the intervention arm and for the control arm. If you are tied on this outcome, you move to the duration of hospital stay and so on and so forth. At the end of all of this, you would take the total number of wins in the intervention arm, the total number of wins in the control arm, the total wins and losses really, and the total wins by the total losses for the intervention arm is the win ratio. Now a win ratio more than one means that the intervention is doing better. A win ratio less than one means that the intervention is doing worse, and a win ratio which is equal to one means that the intervention is as good or similar to the control arm, as in we have not demonstrated a difference at least. So an additional thing that they did in Andromeda shock is they categorized or stratified the patients by an Apache 2 score less than 19 and more than 19, and they ran those win ratios by both those strata. And I would sort of summarize by saying that ⁓ the overall win ratio is in favor of the intervention. Now in the Apache 2 score less than 19, while the effect estimate is in favor of the intervention, the conference interval is crossing ⁓ the point of no difference. Whereas for Apache 2 score more than 19, the conference interval is fully on the side of favoring the intervention. So overall, Binjay Show is supporting the intervention. ⁓ In higher Apache, there is clear evidence of that. In the lower Apache group, is a cross across the intervention, the conference interval is crossing. but still the point estimate is in favor of the intervention. So this is the strategy that they've chosen. I'll also very briefly talk about the secondary outcomes. The secondary outcomes were the components of the composite outcome. So they looked at mortality as a component, the organ support three days, and the length of stay. Ramesh Venkataraman: Please. Bharath Kumar T V: And mortality was not different between the two arms. was about 25, 26 % in both. there is probably a, the difference is probably coming from the duration of organ support or organ support three days, which seems to be one day higher in the ⁓ intervention arm. And the length of hospital stay is also not different. So this is sort of ⁓ a summary of their results. So again, ⁓ what do you think about this approach, but also about the results themselves? Ramesh Venkataraman: this approach is kind of cool, right? I mean, you have three outcomes and then you weight the outcomes and then if they tie on one outcome, then you move on to the next outcome and so on. So I think it looks like it's a fair way to look at it. However, I think we need to understand a few things like you rightly pointed out when I count all the wins in one side and all the wins in other side, the intervention had more wins. And therefore, we seem to think that this is a positive study. Bharath Kumar T V: Yeah. Ramesh Venkataraman: And then you also nicely brought up the fact that among the three, the main difference was seen in the organ free days, not in the mortality, not in the length of stay. And the intervention group had one day less of organ support compared to the control arm, pretty much. And we don't know how the win is defined. Am I winning because my noradrenaline was shut off one hour earlier than the other group? There could still be a win. The authors don't mention exact criteria based on which a win was given to a particular group, right? If they say, okay, one day difference is a win, less than one day difference is a tie, then at least there is a clinically meaningful difference. But if the difference is only a few hours, then is there a clinical relevance to it? Shutting off a vasopressor a few hours earlier may not be a great win for me in the larger context of things if the mortality is roughly the same. Bharath Kumar T V: Mm-hmm. Ramesh Venkataraman: The second thing if you dwell in the data very closely and you look at that even within the organ ⁓ free days the percentage of patients who required organ support at every time point that was checked was essentially the same between the two groups. So this indirectly tells me that if the proportion of patients requiring organ support was similar across both the groups at various time points may be the difference Bharath Kumar T V: Mm-hmm. Ramesh Venkataraman: in the duration of organ support between the groups was very very small and I don't know the clinical relevance at the bedside for me of this small difference and even though this idea of wind ratio looks cool and the study is positive from a clinical point of view whether this is a meaningful difference is something that I am skeptical of. Bharath Kumar T V: Mm. Yeah, yeah, no, that's a, that's, those are great points. I was also sort of curious and did some more digging in the supplementary appendix. And again, I found that in the organ support three days, if we take vasopressor or ventilation or renal replacement, the difference that I could see from their tables is the mean difference is about 0.6 or 0.7 days for each of this. So obviously, you know, how clinically significant that is, is a very important question. And it all seems to be the whole win is driven by that rather than the mortality, guess. So that's a super important point ⁓ for listeners to note. ⁓ So given all of this, Dr. Ramesh, what do you think are the big messages coming from this trial? ⁓ And how do you see things moving forward? Ramesh Venkataraman: So, again let us go back to the original rationale one, whether a personalized targeted resuscitation than a generic kind of fixed kind of targeted resuscitation makes a difference. Here they have picked physiological variables that you and me would clinically intuitively agree upon. Bharath Kumar T V: Mm. Mm. Ramesh Venkataraman: However, we need to understand that none of these targets have been validated in a prospect of large database or a prospect of larger trials. Yes, I agree that diastolic blood pressure will prompt me to increase noradrenaline, but whether 50 is the cutoff is something that we don't know, we do not validate it. So one, yes, I agree with the fact that we need to personalize the resuscitation to our patients, but I'm not 100 % sure whether I need to follow the exact same targets that these authors Bharath Kumar T V: Mm. Ramesh Venkataraman: at this point of time have basically told us. Having said that, if you ask me what targets would I follow, I mean, I really do not have any targets based on evidence at this point of time. This seems pragmatic, this seems clinical, but I am just throwing that a little bit of salt of skepticism inside basically saying, no, do just get fixated on these particular targets when you resuscitate. This is number one. Number two, answering the second question, CRT seems to be a very easy tool. and a target of resuscitation and maybe we should incorporate it also along with whatever else we do. If you have ACV02, VAC02, lactate, whatever you have, then CRT can also be one more tool that can probably be looked at because it's easy enough. In a resource limited setting, this could be a very, very valuable tool in terms of guiding our resuscitation and whatever protocol you tend to follow would still be. worthwhile looking at CRT as a particular target. And it's interesting that CRT is a target because to do a CRT, you have to go to the bedside, you have to press the pulp of the finger with a slide for 10 seconds, wait for it to blanch and then use a chronometer to see how long it takes to resolve. In that time, I'm sure having a caregiver at the bedside would probably pick up other issues, like suddenly a little bit of tachycardia, a little bit of hypoxemia, whatever it is. the caregiver could pick that up and interventions I'm sure would have been done and that also could have helped in terms of the outcomes rather than just the CRT checking alone. So we need to be able to differentiate whether it's a CRT or just somebody going into the room every hour to check the CRT that changed the outcome. So that's something that we need to be ⁓ careful about. So broadly speaking, I agree that physiologically. specific resuscitation catered to the individual's variables is definitely a way forward. We are not there to say these are all targets that are validated and we need to use. I think CRT is a great tool and I kind of like the fact that you have to get into the room then be a remote physician for you to be able to assess the patient and more caregiver time at the bedside definitely is going to improve outcomes. I'm thrilled that Bharath Kumar T V: Hmm. Ramesh Venkataraman: Now this is probably going to start getting practiced and hopefully will improve the overall standard of care. Bharath Kumar T V: Now, those are all brilliant points again. And I think that the last point that you made was also highlighted in an editorial that accompanied the paper, which also said that it's probably the clinician at the bedside and the repeated assessments that makes a difference. So I want to also talk about a couple of maybe other points that I wanted to bring about. One is, of course, we obviously want to congratulate the investigators for a phenomenal trial. ⁓ Despite all of the limitations, et cetera, this is a very hard trial to do. It's done across several countries. The adherence to the protocol is excellent. And people have often said, we want to do a personalized strategy. But ⁓ these guys are actually testing it, right? They're actually testing a personalized strategy. So I think this is a remarkable advance as far as ⁓ at least testing interventions goes. So it's a fantastic effort. And I think it's a trial with a lot of internal validity. The question really is about the, I guess, the external validity, the generalizability. And the other point that I wanted to bring up is This is a resource intensive protocol. So if this protocol is going to bring about not a difference in mortality, but a marginal difference in organ support, whether teams across different settings will be able to implement the whole protocol, ⁓ how resource intensive it is, is obviously a question that comes up. But I agree with you that it's more assessments, clinician at the bedside is probably what. ultimately matters. ⁓ So I guess ⁓ on that note of clinicians at the bedside, more assessments and sort of directing care, we will stop this episode. We want to thank everyone that's joined in to listen. Thank you so much for joining us. And we hope to keep the series going. Thank you so much, Dr. Ramesh. Ramesh Venkataraman: Yeah, thank you, but it was fun and I hope to do more sessions again. So look forward to doing it. Yeah, thank you. Bharath Kumar T V: Thank you.