Good morning, everybody. Welcome to this special session devoted to the topic of sepsis, protocolized versus personalized management of sepsis. My name is Robert Stephens. I'm from Johns Hopkins University. And it's my great pleasure to introduce my co-chair, Dr. Samantha Gambles-Farr. And without further ado, we're going to start with our first speaker, Dr. Ricard Ferrer from Valencia. So, thank you, thank you for the organizing committee for inviting me. I am Ricard Ferrer from Barcelona in Spain and so this is not my first slide. Okay, so I'm going to discuss about protocolized medicine in sepsis. So, the idea is what about variability in care if this is good or bad. And I think that, yes, that there is variability in the treatment of sepsis, of course. This is data coming from the IMPRESS study, it's a point prevalence study done a few years ago, including patients in 62 countries. And you can see here that bundle compliance was different from one area of the world to other, also for this six-hour bundle, so there is variability in bundle care and of course difference in mortality in these different areas but can be explained by many other things, not only by bundle compliance. If we come here to the U.S., this is time to antibiotics, there is variability in each center. Here there are 24 ICUs that look to time to treatment in antibiotics in sepsis, there is variability in each center but also variability center to center. So the message is that there is variability in sepsis care but probably this variability could be accepted if this variability is coming because the autonomy of the doctors applying their knowledge and their experience when they treat patients with sepsis but could be unacceptable if this variability is coming because lack of knowledge or because faulty application of the knowledge. The next question is that physicians perform better if they are just using their clinical knowledge or this performance can be improved with guidelines, protocols and checklists. So protocols per se are not a warranty of better care. This is coming again from the U.S. These are different hospitals with different number of protocols and you can see here that more protocols seems to improve care or at least mortality but higher number of protocols increase mortality again. So protocols per se don't mean an improvement in performance or in mortality. So are protocols feasible in sepsis and do protocols improve clinical performance? So I'm going to go back 40 years later to this study by Roger Born and co-workers where they meet to design the inclusion exclusion criteria of the esteroid trial as I said 40 years ago with the hypothesis that they can reduce or cause mortality using esteroids in sepsis like a magic bullet. We know that the results of these trials were neutral and in fact it was harm in some population of these patients but the important finding of this study is that they can describe a population with different mortality according to severity that were the base for the first sepsis definition. So this is the first time that patients with different infections and severity are collected a group in sepsis. Bill Dellinger was one of the authors of these guidelines, of this definition or criteria for classifying patients, for grouping patients with sepsis. Despite sepsis is very heterogeneous and I'm sure that Dr. Wong will discuss about this. You have different infections, different host, different microorganisms, different environment, different microbiome and probably we can now classify patients about different endotype and maybe we can classify patients now by different phenotype. There is a lot of heterogeneity despite that I think that it is still useful to group patients in sepsis to have this group approach. Of course if there is heterogeneity there is heterogeneity also in treatment effects. So for example in this simulation some patients are going to be, this treatment is going to be beneficial in some patients, it's going to be neutral in another group of patients and it's going to be harmful for some patients. The result of this trial of course is going to be mainly neutral because it's not a clear benefit of harm and this is a consequence of the heterogeneity of sepsis and this is the reason why the main trials in sepsis has been negative. But if we have the comparison with cancer you can see here that cancer has a more personalized approach and they have moved from cancer like sepsis to each one of the diagnosis, some of them with molecular diagnosis and with this personal approach they were able to increase survival of patients but again with well established protocol for each symbol condition. So a more personalized approach is not against the use of protocols. So why keep lumping or why keep grouping patients? Because some of the strategies are common for most of septic patients. For example antibiotics, this is an 18,000 patient study we published several years ago showing that there is a relationship between time to antibiotics and mortality and this has been reproduced in many other studies, in this case from Northern California with 30,000 patients. And again this relationship between time to antibiotics and mortality and also from the New York state data showing that there is again a relationship between time to antibiotics and mortality. So this is common for patients with sepsis and septic shock and this is playing part of the successful of this grouping that has been done 40 years ago. Also this is true for source control, early source control, lower mortality. But again if you look for source control there is a lot of heterogeneity that we have to try to reduce. These are trials for antibiotics and you can see here the amount of patients that receive inadequate or late source control. So again in one trial it's 5% but in another trial 30% of the patients receive inadequate or late source control. So again a lot of variability that maybe can be reduced using protocols or checklists. So you have seen the difficult that is to generate evidence in a condition that is not a condition, it's a syndrome with a lot of heterogeneity. Difficult to have randomized control trials in this setting. But we have some strategies that despite this evidence is coming from many observational studies we are able to condensate this evidence in the guidelines that are produced by the Surveillance Sepsis Campaign every four years. And bundles of protocols are just tools to implement this knowledge, this evidence that is condensated in guidelines. So mainly protocols are tools for better clinical performance. So the usefulness of the first hour bundle is that for every emergency department, every institution you can be sure that at least the basic things are done or at least started, which is helping to reduce some of the variability and to increase compliance with some of the recommendations based in evidence. There are some intervention studies trying to demonstrate that this bundle approach or this protocol approach is useful. This is the intervention we've done many years ago in Spain, it's a before and after design. And after the intervention based in bundle care we were able to show a reduction in mortality. To repeat this intervention some years later, just focus on early antibiotics and we were able to reduce time to antibiotics significantly without a reduction in mortality. Maybe because the benefit is always coming from a bundle approach and not for a single element of the bundles or maybe because the study was under power to show that. Also in the pediatric population we did this intervention, again focus on early antibiotic in kids with sepsis and there is a reduction in time to antibiotic again with an absolute reduction of three points in mortality in the pediatric population but it was not significant again maybe because the study was under power. There are studies with a better design, in this case is a cluster randomized design with some hospitals randomized to usual care and others randomized to bundle care and protocol care and there is a reduction in mortality of 2% only in the intervention group showing with this night design that is feasible to see this reduction in mortality coming from this bundle care. And also this is data again from New York State showing the increase in compliance with bundles and decrease in mortality in each period with this approach. So the question now is this approach of grouping patients with sepsis, is it still valid? I think yes. There are some questions. You know that there are some political issues. Not all the societies are supporting the swine sepsis guidelines. I'm not sure that this discussion is mainly technical, I think it's more political. There is of course some guidelines misuse, so guidelines are designed to guide our care but cannot be transformed in a paper performance for example. And of course there are these editorials talking about sepsis hysteria saying that we are over treating patients because the use of these approaches. But on the other side we have many things that can be improved by this LAM approach. Sepsis awareness, there are some patients still coming to our hospitals with a lot of delay and we need to work on that. There is, as you have seen this morning in the opening session, there is opportunities to early detection of sepsis using artificial intelligence approach or other approaches. There is still room for improvement for interventions that try to improve the process of care inside of the hospital, for example using fast track or sepsis codes. All the innovation in the microbiology lab coming from rapid diagnostic tests can be incorporated in the process of care of patients with sepsis and of course all the techniques which are less invasive for source control should be incorporated in our process of care. This is some examples from my country, this is the regional sepsis code and you can see here how patients are transferred to different hospitals using protocols for transferring patients to another hospital, to a referral hospital with the possibility of doing any type of source control or any type of ICU support according a protocol. This is the protocol of my hospital and you can see here how patients with severe sepsis or sepsis have a protocol for resuscitation, another one for sampling, another one for infection control and another one for source control and of course we have protocols for each of these parts of the process like empirical antibiotics, they need of course a specific protocol. This is how the process of care is in my hospital, so it's a multidisciplinary process of care, we have several steps here, several quality indicators that help us to understand if the process of care is done well or not. So with that my conclusions would be that despite sepsis is a very heterogeneous syndrome with important variability in terms of infections, host, etc., despite that we can organize the care of this condition using protocols. I cannot imagine any organization, any scientific society, any hospital that do not work with a standard operating procedure, so we need that to reduce variability and to improve the process of care. We have the example of cancer as I mentioned to you, it's a broad concept but when you go to a specific conditions and personalized care you have to use protocols for doing that. So I think that in general personalized care is not against protocol management, it's something that needs to be combined of course. Thank you for your attention. Thank you Dr. Rocha, we're going to take questions after each speaker. There are microphones out in the audience, are there any questions for Dr. Rocha at this time? May I ask a question? You know what, I'm working on this precision medicine and the protocolized medicine but I'm looking into it from a different perspective and that's the barrier to education. What I have found that with all this protocolized medicine has been so much now upgraded into epic system and electronic medical records and everywhere, now I see that residents, fellows, house staff, whenever they are starting to treat the patient, they just go and click, click, click, click and they think they have started and I see this as a barrier to education. I feel that protocol is needed but I think we have gone too far on the pendulum and every patient is different. For example, norepinephrine, I start norepinephrine, patient is not responding by six, eight mics, I start vasopressin, boom response but I get sometimes this argument that vasopressin is not a part of the protocol, it needs to be norepinephrine, norepinephrine is the first line but I know that in this patient, norepinephrine is not working, vasopressin is working or if I put my probe, ultrasound probe and I see that RV is a little down or LV is a little down and I jump on dobutamine early. So I think protocol is needed as you said but I think we have gone other part of the pendulum, it is causing a barrier to the medical education and we need to, rather than precision medicine, I will use the word personalized medicine for each patient. We have to, we can't put all the patients in one port, that's what I'm giving my opinion, that's what I have learned from my clinical practice. Yeah, I completely agree with your opinion, I mean that guidelines should be adapted in local protocols and then we have our autonomy to decide in some topics what to do at bedside, of course, and your clinical expertise and your knowledge are going to be key in the management of your patients, of course, but it should be based in knowledge and clinical experience. Thank you for that great talk. My question is, you made a statement that protocols are tools. I think the intensivists who really work in most of the ICUs will tell you that they disagree with you more, the way the medical legal stuff works and the way the clinical decision making works is horrendous here and I'll tell you why. The guidelines say that you have to give 30 cc's per kg body weight to every patient, doesn't matter if you are in a CHF, doesn't matter if you are on a congestive heart failure, doesn't matter if you are on dialysis or whatever and there could be multiple but you have to give it otherwise you will be reported and you won't be paid for it. Second, you say that these things are, and all these things are good, you did not mention any study where they said earlier the antibiotics, you told us that there are many antibiotics which are, I'm sorry, the earlier the antibiotic the better the outcomes are. There's two studies that have come out that say actually if you give antibiotics in the ambulance, that was in American Journal of Medicine, the mortality may actually go up. So earlier it is not always better, yes, most of the time it is better. The more fluids are not always better, yes, they can be better in some time. By the way, this coronavirus that has come in, the WHO has said that you have to give 30 cc's per kg body weight to everyone. So these are not just tools, you are telling people what to do and that is actually killing people and we do this every day in real ICUs. We don't do this in isolated places where guidelines are created. So please be cognizant of everything. This is not silos. This is real patients and you have no idea. Every study, all these studies that are coming out with the fluids, nobody, SCCF still wants us to give that much of fluid. It's been taken over, not fair. So thank you for your comments. I'm going to discuss about two area of comments. The first one is about antibiotics. You mentioned the study about pre-hospital antibiotics. If you do a randomised control trial, you need two things. First, if you want, you need prognostic enrichment. That means that the outcome, which is mortality, should be frequent and this study is 5%. So you cannot design a study about mortality with antibiotic with the outcome is only 5%. The second thing you need is that between the intervention and the control group, there is some difference. In this study, the patients in the ambulance received the antibiotics few minutes before that the patients that received the antibiotics in the hospital. That means that between two arms, the difference is only minutes. It's impossible to show difference. So the results of this study, this is a well, nice study, but we cannot draw consequences about early antibiotics based in this study. So I do believe that early antibiotics are very important for all patients with sepsis and septic shock and data for this study, I'm not changing my mind at all. Second point is about early fluid resuscitation. If you read the statement, it's begin fluid resuscitation during the first hour. Don't say that you have to give 30 ml per kilo during one hour. That means that some patients, of course, do not need 30 ml per kilo. Some patients cannot tolerate 30 ml per kilo, of course, but this is your clinical expertise and your behave at bedside that have to guide your therapy in those patients. But of course, for antibiotics, you need a local protocol. You cannot use guidelines, any guideline for any society to guide your antibiotic therapy. You need a local guideline adapted to your local ecology. And I think most of you agree that you need a protocol for that because most of the time, the one who is going to be at bedside is going to be a junior physician that is not able to understand the local ecology, the pharmacokinetics and pharmacodynamics of each antibiotic in a patient in septic shock. You need a guideline or a protocol in this case for doing that. Is there time for one more question or? We'll try to get to your question at the end, I'm sorry. For the sake of our speakers, we'll have to move on to our next speaker. At this time, I'd like to welcome our next speaker, Dr. Wong from Cincinnati Children's Hospital Medical Center. All right then, let's talk about something more or less controversial here. Very good. So thank you for having me here. So these are my disclosures, a few grants from the NIH, a number of sepsis biomarker-related patents that I hold with my institution, and yes, I will be speaking about them, and then some positions on some advisory boards. So this session has been sort of framed as one versus the other, and I don't think of it that way. I think protocols are fine. I think that they're necessary. I think they're important. But I would also say that our patients deserve that we aspire to more. They need us to aspire to think bigger, to think beyond protocols and bundles, and to do more and to aspire more. So one of those aspirations is this concept of precision medicine, which fundamentally means that you simply align therapy with the underlying biology and the risk of the individual patient, right? Not very earth-shattering, but that's the fundamental concept there. And when we think of precision medicine, most of us think about DNA and how gene variants can influence susceptibility in response to therapy, but there's much more to it than that. We also think about it mainly in oncology, and certainly in the oncology field, that's where most of the progress so far or the advances have been made with respect to precision medicine. But I would like to suggest to you that critical care is very amenable to this, right? If you think about the patients that we look after, we're not looking at specific diseases that are monogenic or anything. We're looking at syndromes such as septic shock, ARDS, and so forth, and as we said before, as Dr. Roca said, there's tremendous heterogeneity there, and we don't really account for that very well. All right? So the challenge here is how do we take these heterogeneous cohorts and subgroup them into sort of more homogeneous subgroups? How do we get there so that we can, again, align therapy to risk and underlying biology? How do we get there? And so to me, the fundamental concept is this concept of enrichment. And so broadly speaking, enrichment means a selection of a population using any characteristic in which your intervention is more likely to be effective when compared to an unselected population. More specifically, prognostic enrichment means that you select a population that's more likely to have a disease-related event, for example, mortality. More challenging is this concept of predictive enrichment in which you select a patient population in which your intervention is more likely to be effective based on underlying biology. It's harder to get at that. And so if we get back to this cartoon, I can start to envision how you can use prognostic enrichment to break up patients broadly into those who are low-risk and those who are higher risk for mortality, for example, in sepsis. The low-risk patients should be treated with your bundles and your protocols. They are going to do well as long as one adheres to those protocols and actually does the things that one is supposed to do. And they're not going to benefit from any experimental therapy. And then you take your high-risk patients and then can you use some form of predictive enrichment to better align your new therapy with their underlying mechanism. And we already do some of this, right? We select our sickest patients for ECMO, right? That's a form of prognostic enrichment. We select antibiotics based on, as we were just discussing, based on microbiology and susceptibility and so forth. That's predictive enrichment. We should select our cardiovascular agents, like we were just discussing, based on the underlying physiology of the patient, right? So if norepinephrine is not working in somebody who has myocardial dysfunction, that makes a lot of sense, right? So we already do this. The MONARCHS trial many, many years ago selected patients for an anti-TNF therapy based on IL-6 levels. The EUPHRATES trial that Dr. Dellinger ran selected patients for hemoabsorption of endotoxin. They were selected based on having a certain level of endotoxin, right? That makes a lot of sense. That's a form of predictive enrichment. I think we can do better. So one of the ways we can do better is this thing that we've developed called PERSEVERE, which is a biomarker-based risk model. It's serum proteins, very easy to do. It's based on classification and regression tree methodology. And it allows us to reliably estimate baseline risk of mortality in kids with septic shock. It's a form of prognostic enrichment. There was a data science session just before this that talked about that your model should be interpretable, right? It should be something that a clinician can look at and say, OK, that makes sense to me. I would argue that this model meets those criteria. It's just a binary series of decisions. And patients are allocated to these nodes based on biomarker cutoffs. Patients that are allocated to those nodes have a very low risk of mortality. Patients there have an intermediate risk, and those patients have a high risk. And very recently, we applied this a priori. In other words, we didn't recalibrate the model. We applied those model rules to a new cohort of almost 500 patients, and it performed very well. And we can talk about that performance later on. This is another way to look at that model, OK? These are patients lumped into high, low, and intermediate risk, and it separates the patients very well. Now imagine the patients in that light blue line that are low risk, right, that you can identify when they come into the ICU. Those patients should not be getting any form of experimental therapy, because you're not going to move the curve on those patients, right? Their mortality risk is very low with standard care. With good, excellent, standard, protocolized care, it's fine. And so experimental therapy is not going to help those patients. It's the other patients that you want to get after. So that's the concept of prognostic enrichment. As we said, predictive enrichment is harder. It's trying to identify subpopulations that may respond better to your therapy based on biology. And this is where this concept of endotypes comes in, which are subclasses of diseases defined by mechanisms or underlying biology. And so five years ago, we reported these sort of two endotypes of pediatric septic shock that we call endotype A and B. We're going to talk more about that tomorrow. They're defined by 100 genes that are expressed with those expression mosaics right there. And those genes correspond to adaptive immunity and glucocorticoid receptor signaling. And as you can see by the coloring there, these genes are repressed in the kids that are endotype A. So because of that underlying biology, we think this can be a predictive enrichment tool. The kids that are allocated to endotype A have greater organ failure and higher mortality. If you account for underlying illness severity, age, and so forth, allocation to endotype A is independently associated with worse outcomes. And what's more interesting from a predictive enrichment strategy is that if those kids get corticosteroids, their risk of mortality is much higher. If you know these data, you know that they're based on observational data, so they have to be interpreted cautiously. But recently what we did is we downloaded data from the VANISH trial. These are adults with septic shock who were in a randomized trial with vasopressin, norepi, and hydrocortisone, and so forth. But a subset had transcriptomic data that are publicly available. We downloaded that data and assigned them to our pediatric endotypes A and B and found the same thing. The patients who were exposed to hydrocortisone had a much higher mortality compared to those who were not. Okay, so external corroboration of what we previously reported in a setting where the hydrocortisone exposure is much more protocolized. So if this is true, it's no wonder there's so much controversy about corticosteroids. This is identifying a subgroup who should not be getting corticosteroids. What about who should get corticosteroids? So what we did recently is we used this prognostic and predictive enrichment to say, can we find kids who actually benefit from corticosteroids? And the answer is yes. If you're endotype A and you have an intermediate to high baseline risk of mortality, you have a more than tenfold decrease in your odds of poor outcome. Again, this is all based on observational data, but it's that concept of combining predictive and prognostic enrichment to be able to identify subgroups who might benefit. And so we're testing this now prospectively. There's a trial that was just launched called SHIPS, which is a hydrocortisone trial in kids with septic shock, and we're obtaining biological samples to be able to do this prognostic and predictive enrichment. Unfortunately, for a number of reasons, that tool is not going to be used to inform enrollment, but we have designed the next best thing is that we have a pre-specified post hoc analysis based on this enrichment. So this will be a chance or an opportunity to test these concepts in a prospective trial. All right, this is meant to be a break. So talk about breaking protocol, right? So that's me, a pediatric intensivist in the woods on an ATV with no helmet, okay? So that's getting a little outside of the protocol. It's also meant to say, as a break, that what I'm about to show you to you may, for some of you, may seem incongruent and a bit crazy, but we're excited about it. So it comes from this paper that was just recently published, and what we did is we tested these biomarkers in mice. And so it turns out that in a mouse model of sepsis in which they get antibiotics and fluid and things like that, these same biomarkers work, all right? So we can actually predict mortality in experimental mice using the same biomarkers, which suggests that there's some conservation there of the biology. And so again, we can, using the same kind of idea, we can break down the mice into low and high risk with a good ARUC. And so we started to characterize these mice, and what we found, sorry, going too fast here, is that the high risk mice have a greater degree of systemic and organ inflammation, and they also have a greater bacterial burden, okay? So what does this mean? Well, it could mean that these mice that are high risk have excessive or ineffective inflammation and can't clear bacteria, or it could mean that they have greater bacterial burden and so they have persistent inflammation. So if this is the case, then perhaps corticosteroids could rescue the mice, but they don't, okay? So if you look at just the high risk mice, you give them an anti-inflammatory dose of corticosteroids, we can't affect their outcome. Which leaves us with this option as well. Maybe if they have greater bacterial burden, we can match antibiotic dosing to that higher burden, so we simply just doubled the antibiotics, and you can see you can start to move the needle. So in the high risk mice, if you match antibiotic exposure to bacterial burden, you can start to rescue the mice. I know these are just mice. So we went back to our data and said, well, this is kind of interesting, do we have any evidence that kids who are higher risk, do we have any evidence that they have a greater bacterial burden? Right? That's kind of hard to do with clinical data, but what we did is we just simply looked at positive cultures here, right? So on the X-axis there are the categories of baseline mortality risk, right, low, intermediate, and high. The Y-axis is the proportion of kids who had positive cultures. And you can see that with increasing risk, those kids had higher incidence of positive cultures. It's not direct evidence of a greater bacterial burden, but I think we can all agree that in order for cultures to be positive, you have to have a certain load of bacteria for them to be positive. And then we looked at the kids who had positive cultures, and we said, okay, how many of those have positive blood cultures, right, because those were – the previous slide was cultures from all over. How many had positive blood cultures? And again, you can see that with increasing baseline risk, almost 70 percent of the kids who were high risk had positive blood cultures, which I think is pretty reasonable evidence, indirect albeit, that these kids may have a greater bacterial burden. All right, so we've talked about – we always talk about early antibiotics, and we talk a little bit about broad-spectrum antibiotics, but we never really talk about matching antibiotic dosing to that patient. And the reality is, is that antibiotic pharmacokinetics in our patients is widely variable. One of our fellows is doing a study right now looking at ceftriaxone in kids in the ICU, and with standard dosing, which is what we all use, with standard dosing, blood concentrations of ceftriaxone vary as much as 40-fold, 40-fold. That means that a lot of kids, and I'm sure a lot of adults also, are getting blood concentrations that aren't sufficient to kill bacteria, right? So what I'm imagining is this concept, is that can we select high-risk kids, right, based on Persevere, low-risk kids, standard care, standard antibiotics, they're going to do fine. But the high-risk kids that have a higher bacterial burden, perhaps those kids should get selected for very intensive PK and pharmacodynamics modeling and measurements so that we can match antibiotics exposure to bacterial burden. And now we have a way of detecting and predicting those kids. We actually just put in a big grant to do that, and we'll see if we'll be able to actually get that done. But imagine if a solution to sepsis is just dosing our antibiotics more intelligently. That's pretty pedestrian, but boy, at the end of the day, it is an infectious disease, right? And we chase all of these other things like anti-TNF and anti-this and that, and perhaps we're not paying enough attention to how well or how accurately or how precise we dose our antibiotics. So again, those are the key concepts. I'm happy to send these slides to anybody if you email me. I guess there's some sort of hub here at SCCM. I can upload it if you will, but feel free to email me, and I can send them to you, including the picture on the ATV. Thank you very much for your attention. Thank you very much, Dr. Wong. We have a few minutes for questions. Thank you for that intriguing, very nice talk. A question. The thing is that I think we are still working in silos, although the precision medicine is going where it is, and I think it's awesome. However, having said that, have you ever thought that maybe the steroids aren't working because the receptors are not ready, because something is missing? Now, I'm going to say something which is very controversial, but this is just to give you an example. Maybe you need vitamin C. Maybe you need thiamine. Maybe you need albumin molecules to actually go to the endothelial endocalyx to be stabilized. We are thinking only that maybe steroids aren't working, and that's ditch the steroids, or for that matter, ditch whatever is not working, and then move on to something else, the magic bullet. Have you ever thought that there might be something else that might be going on at a mitochondrial level, which might be missing? Maybe it may be zinc or magnesium, anything like that, and the receptors might not be ready enough for that yet. Do you have any comments on that? Yeah, actually, we've actually published data that, well, thank you for the question. I mean, we could talk about this all day. That was just a long question, a lot of points. Thank you. So, we actually have data showing exactly what you said, that the receptor for glucocorticoid is very altered in kids, and there's some data in adults also. So, yeah, there is a problem at the level of the receptor. We think that this tool is a way to broadly find responders and non-responders, but I'm sure that there's finer granularity that one could get to. I have no doubt about that. I guess my point is that's the way we need to think, right? We can't just keep doing steroid trials and just pile in 1,000 patients and expect that something good is going to come at the end, right? I think we need to account for that underlying biology, whether it be mitochondria, whether it be zinc, whether it be receptor levels, we need to find a way to type these patients. At the same time, you need to be pragmatic, right? Because in trying to do this in a clinical trial, you need to get data back very, very quickly, right? That's the disadvantage that we have compared to the oncologist. You can go to an oncologist, and it can take two, three, four weeks to genotype your tumor, to find all kinds of markers, all kinds of details. We don't have that time, right? We need to have that kind of phenotyping or genotyping within one or two hours in order to make decisions. So you need to balance the level of detail with that reality, that decision-making in our patients is very time-limited. But yes, I could not, in principle, agree more with you that there's more to the story than just what we're showing. Just to follow up, there are studies that the steroids that we use in ARDS, we have to really specifically, I'm not going to name names, and in sepsis, there have been different hydrocortisone versus whatever you want to call that. There are different studies showing that one is better than the other. Even the steroids themselves are not equivalent. Do you agree on that? And I think that is because there are different receptors in different pathologies. And that is why, you know, we are lumping everything together, and that's where the issues come. Yeah, I agree. I mean, I think that there, biologically, to me, it makes sense that steroids can work in ARDS and sepsis. The challenge is finding which patients, what's the underlying biology in which that's going to be? So that's what precision medicine is about, is finding that. We're a bit concerned, is because we're not just finding who should get steroids, but we're also identifying a group that should not get steroids. And if that's true, we have a big problem. And I think it behooves us to be able to find those patients, at the very least, to avoid corticosteroids in that population. We're not doing that. We're not doing that. And so I worry that we are hurting some patients with some of these protocols, because we're not finding those who could potentially be harmed. And like I said, I think we need to aspire to more, and patients deserve it. Sure. So, Dr. Wong, your research has focused a lot on the diversity of the host response, but what about the diversity of the infecting pathogens? And I was wondering if you could comment on any recent research on rapid microbiological molecular assays so that we can quickly identify which is the pathogen and target selectively the antibiotic. Not just the question of dosing the antibiotic, but also selective targeting based on the pathogen. Yeah. I mean, I think that's a whole, like I said, after all, it is an infectious disease, right? And in our studies, we haven't found much of a microorganism-based signal, remarkably. I'm still not sure that I believe that, but we just haven't found it. But yes, there's a lot of tools coming online now for more rapid detection, not just of organism and class, but also of antibiotic resistance, and I think that could help us a lot more. Again, we need to go further, and I think we need to understand the microbiome much more, because that plays a huge role in what we do, and it's untapped. I have a colleague who has a really interesting idea, in that we always talk about source control, but when we talk about source control, we talk about the abscess here, or the abscess there, or some other clear source of infection. He has this theory where the gut microbiome may be a place where we don't have source control, right? So that you may have Klebsiella sepsis, and you're using antibiotics for that, and you're clearing the blood, but the gut is still colonized, and still populating, and still translocating, and so you don't have source control. So we're actually starting a clinical protocol now, or a clinical study, to start testing that hypothesis, where we can do shotgun sequencing of the stool microbiome to test that. It's a really interesting concept, because we always think about source control, and it's really important, but his idea is the gut is potentially a source that we're not controlling or accounting for. I think it's fascinating, and I guess it's what's going to keep me around for another four or five years to do stuff like that. So thank you for that. Thank you very much. So it's my honor to introduce our third and last speaker, Dr. Philip Dellinger from Cooper University Hospital. Thank you very much. It's a pleasure to be here. The title of my presentation, Management of Sepsis, Protocolized and Precision Approaches, are both needed. That's not a good start. Right click, left click to advance. I'm left-clicking away. So I've broken my brief talk into three portions, some comments about protocols, some comments about precision medicine, and a summary on where I think we stand in 2020. Relative to these two approaches, I have no conflicts of interest concerning protocols. One of the big advantages of protocols is it's very useful for those less initiated to the cure of the disease process at hand. So someone that doesn't see severe sepsis on a regular basis may very much benefit from protocols. But remember that even highly trained pilots still rely on checklists to make sure that they have not missed an important component. Protocols also allow scientific rigor to be applied to test the value of the protocol. This is mostly pre and post. The surviving sepsis campaign, sepsis bundles, arose in 2005 and as part of the International Performance Improvement Program, allowed seven and a half years of data to be analyzed, looking at how performance on the sepsis bundles changed and how it might be associated with mortality. Education, as part of this Volunteer Performance Improvement Program, was web-based, face-to-face. Regional and national and international meetings and audit and feedback, these were associated with protocol implementation. At the participating hospitals, you can see that there was an unadjusted and shown here, risk-adjusted decrease in mortality. By quarters for those hospitals participating, meaning that hospitals that were in for eight quarters, you'd look at how process improvement and mortality changed over that time. So clearly, an association between use of protocols over time and better outcome. Rory Staunton had an unfortunate death as a youngster in New York, leading to a state mandate that all hospitals in the state of New York would participate in a formalized performance improvement program. The data from that performance improvement program was published showing which required the inclusion of what was the surviving sepsis campaign three-hour and six-hour bundles. And each hospital had to have a protocol. You can see that over time, compliance with the protocol three- and six-hour bundle increased and mortality decreased over time. And here more recently, again looking at the mandated New York sepsis performance improvement program and the three- and six-hour bundles comparing the New York mortality over time with surrounding states, comparing 2011 to 2013 with 2013 over 2015, which is before and after the mandatory sepsis performance improvement was initiated over a million sepsis patients. And there was a greater decrease in sepsis mortality in New York compared to control states that did not implement sepsis regulations. The limitations of looking at sepsis protocol data, it tends to typically be historical data pre and post, which is not the highest level of evidence. The other issue is that when you initiate a sepsis performance improvement program and a sepsis protocol in your hospital, there is obviously a culture change. And just the culture of awareness may create more attentive care and may alter and move the needle toward more positive outcomes. Well, what about precision medicine? You'll hear the term personalized medicine used to imply what many call precision medicine. This is more of an older term. I think Obama popularized personalized medicine. The problem with using personalized medicine is the implication in some people's minds may be that it implies that drugs or medical devices may be unique to a single patient. When in actuality, a better term is precision medicine, defined as tailoring of medical treatment based on common characteristics for a particular group of patients. Probably the first element of precision medicine would be taking a general clinical diagnosis and making it more precise to a particular subgroup of that clinical diagnosis. This is one example with fluid resuscitation of sepsis-induced hypotension. What about patients with sepsis who have end-stage renal disease on dialysis or compensated congestive heart failure? Should their clinical approach be outside of the protocol for the 30 mLs per kg crystalloid as part of the initial fluid bolus? Chalkboard logic would have it that perhaps there should be concern in those patients about 30 mLs per kg of fluid. However, there's also rationale for why that particular groups of patients would have the same amount of leak and venodilatation and need the same amount of fluid. There is data to indicate that patients who are being resuscitated for sepsis-induced hypotension as part of protocolization that get fluid resuscitation, if you look at the PaO2-FiO2 ratio over time as shown here, that there is no difference in change in oxygenation over time with the amount of fluid that's given to patients with a history of heart failure versus no history of heart failure. In fact, if you look at these two plots, you see that, if anything, there is less of a decline in PaO2-FiO2 ratio in the patients with a history of heart failure. And again, recent data looking at intubation related to fluid resuscitation in patients with sepsis-induced hypotension who have heart failure end-stage renal disease or cirrhosis using a propensity matching comparing 104 versus 104 propensity match with either standard resuscitation, 30 mLs per kg crystalloid or more, or restricted resuscitation less than 30 mLs per kg. And again, you see no difference in probability of intubation within 72 hours. In fact, the curves would appear that there are less patients being intubated that get the higher fluid dose. And finally, this is Kaiser Permanente started a new version of their sepsis performance improvement where they used the three and the six-hour bundle in patients that had lactates between two and four and not just the four or more. And when they introduced this new enrollment criteria in the system, there was hesitation again about the 30 mLs per kg crystalloid bolus in patients with heart failure and end-stage renal disease. And it had, it was somewhat of a hard sell, but in the end there was compliance. And I'll have you look at this, the top red brackets is the change over time pre and post bundle on hospital and 30-day mortality. And you can see that there was a significant decrease in mortality in this group of patients. And if you look at the second red brackets, you see that the benefit was almost entirely in patients that had history of heart failure or end-stage renal disease. So moving on to another precision marker already covered to some degree by Professor Wong, a use of vial markers for precision. This is a publication from 2014 that I was a co-author pointing to trying to deal with the heterogeneity of sepsis by using mediator-based target or what some people would call theragnostics. Professor Wong alluded to this trial that I participated in which we only enroll septic shock patients with high endotoxin activity in the blood as a theragnostic approach. And we used a endotoxin removal device, therefore the second piece of theragnostics. Unfortunately, we did not show any survival benefit, although there were subgroup analysis that have physiologic basis where there were better outcome with the endotoxin removal. What about precision using genetic makeup? Again, Professor Wong discusses somewhat a couple of examples. This is one study using transcriptomic analysis of peripheral blood leukocytes in the UK genomic advances in sepsis study in 29 ICUs, 270 patients, in which an SRS1 signature cohort was identified as an immunosuppressed phenotype that may have import as we move into clinical studies of immune stimulatory therapy and being able to better identify using precision medicine to better identify groups of patients that might benefit from immunostimulatory therapy. Phenotypes and machine learning is another area of precision medicine applicable to sepsis. This is a very interesting study from 2014 looking at patient data from two ARDSnet trials, ARMA and alveoli that identified two phenotypes, phenotype one with less inflammation, shock, and metabolic acidosis that had better clinical outcome and did better with low PEEP in the alveoli trial and phenotype two with more severe inflammation, shock, and metabolic acidosis with the worst clinical outcome, but that did better with high PEEP in the alveoli trial, giving you a sense of what you can do to potentially identify patients that might be targeted for a particular alternative of two therapeutic approaches. And this is just a, shows you a wide variety of clinical and laboratory and physiologic variables that were used, showing on the left separation between the two variables, between the two phenotypes, moving from one phenotype being higher on the left of the graph and then being lower on the right side of the graph and in the middle of the graph not much difference or no difference between variables that were used in the transcriptomic analysis. And this is another recent study published in late 2019, used a derivation cohort from observational database to look at creating phenotypes and then validating whether those phenotypes performed in a larger validation cohort made up of clinical trials in sepsis, the observational data was from the Seneca project. And the four phenotypes identified not only different host response patterns, but also were able to predict different clinical outcomes when retrospectively applied to clinical trials. And finally, this is just something that we're involved with at Cooper. We're involved with a company called Desena, and I guess this is a disclosure, I received no financial support, but I am on a grant application looking at the potential to use machine learning. And machine learning is when the computerized software system, the artificial intelligence is able to actually learn over time and move forward in honing the skills either for identification or for treatment without reprogramming. And what we're looking at is based on what's being studied in the Clovers trial would be the ability of machine learning to identify septic shock phenotypes that would either do better with more fluids and less vasopressors or more vasopressors and less fluids to maintain mean arterial pressure. So I'll finish with a summary slide on where we stand. I think protocols are still the go-to treatment approach for the less experienced bedside practitioner. I think for the season-niched bedside provider that has a lot of experience in severe sepsis, a more nuanced approach based on that knowledge and experience with the protocol is back up. And finally, the future is likely to be with phenotypes, genotypes, and machine learning through either big data analysis or through artificial intelligence and machine learning. Thank you. Dr. Dellinger, thank you very much. Are there any questions from the floor quickly? All right, Dr. Dellinger, let me take you to the point when you were in 2001. You were involved in a study called Zygris. And I heard you when I was a young attending staunchly defending how Zygris was the greatest drug ever to take care of sepsis. And then, by the way, you mentioned that in 2005 this surviving sepsis came out. That SCCM, that was fully funded by Lilly. The Zygris belonged to Lilly. And then we found out that later on that the surviving sepsis basically whatever it, what we found out that the, and I really don't want to go there. Having said that, it was tainted. And it was tainted because they were underfunded by a lot of investigators or funded. And again, it was against ethics. And then in 2011, it was taken back. You're now defending the fluids. And three studies have come out, international studies, giving fluids. And SCCM and you are defending it like anything. And I think that it is just a saving face. And all this data that you presented, I work in New York and Department of Health in New York. Most of the hospitals, the data that we send you is tainted. It's tainted because we don't want our hospital's name on the websites. We do not want to be underpaid. You think that this is good? 6,000 people just signed for WHO that we should not be giving so much of fluids to patients with the coronavirus. Respectfully, sir, do you have a question for Dr. Dellinger? Yes, it is a question. Yes, absolutely. The question is coming. For the sake of time, please state your question. Can we please find out from Dr. Dellinger why we have to give 30 cc's to everyone and not, where did these 30 cc's come from? And I know that you're going to tell me those two studies it came from. But is that fair that everybody is going to get it and those studies that you just said are basically been proven wrong, just like the Zygris studies? Yeah, so I'm almost certain I never said that Zygris was the greatest thing ever. I did think that it worked at one time and it was FDA approved and it was initially recommended by the surviving sepsis campaign. I agree that the surviving sepsis campaign. I have the copy if you want to. Can I have the floor briefly? You can show me afterward. I think that the surviving sepsis campaign was tainted initially by industry funding and I think that was corrected over time. I think the 30 ml's per kg fluid will likely stand the test of time as being an approximate amount of fluid that works for the great majority of patients. Even in the crystal trial, patients are getting one to two liters of fluid before randomization to either more pressers, less fluids, or more fluids, less pressers. In the three trials that make up the prism, the process arise and promise. If you look at how much fluid was given prior to randomization to one of the treatment arms, it's 29 ml's per kg with mortalities less than 20% in those studies. I think the real area that we need to find out is following that initial bolus of fluids when we need to continue to maintain mean arterial pressure should we be doing it with more pressers and less fluids or more fluids and less pressers. I don't think it's about that initial right out of the box fluid bolus. Yeah. So, Dr. Dellinger, thank you very much. I'd like to personally thank my co-chair and all three speakers at this session. This brings to a close this session on personalized and protocolized care for sepsis. Thank you very much.

sepsis treatment

protocol variability

IMPRESS study

bundle compliance

mortality rates

personalized medicine

biomarkers

genetic analysis

precision medicine in sepsis

machine learning

Sepsis

Research

Guidelines

Infection

Presentation

Select

Professional

Associate

2020

Foundational

Sepsis Resources