Good afternoon, everybody. My name is Marlies Ostermann, and I am delighted to chair this afternoon's session. I'm delighted to be in very good company here with my three fellow speakers. And in the next hour, we're going to explore some evidence gaps, gaps in our knowledge related to sepsis management. I will start with a short presentation related to screening and identifying patients. So as we're waiting for my slides, here we go. I practice in London in the UK, and I was very fortunate to be a member of the surviving sepsis guideline, which was published in 2021. These are my disclosures, and they've not affected what I'm going to say next. So as you all know, and I'm sure you agree with me, the surviving sepsis guideline published in 2021 was a major achievement. It was the product of work which had been going on for more than two years, where a large panel of international experts reviewed the evidence and updated existing guidance and recommendations. And they also looked at new areas and added new guidelines. And screening was one of the additional guidelines which had not been included in the previous guideline. And just as a reminder, this is what the guideline says, this is what the 2021 surviving sepsis guideline says. Hospitals and health systems are required to have a performance improvement program in place to improve the recognition of patients with sepsis and their management. It was considered to be a strong recommendation. And in addition, the panel felt that the QSOFA score could not be used as a tool to identify patients with sepsis. It performed in the recommendation. They recommended against this score compared to traditional scores like SIRS and NEWS or modified warning scores. And then lastly, they said once patients are identified, then patients should have a lactate value measured. So these are new guidelines which appear in the guidelines document for the first time. And clearly, they're very important. And as I said, I was part of the panel and I fully support them. And I'll just give you some bit of background, how they were developed and why the panel felt strongly. So the recommendation to have a system in place in hospitals and healthcare systems is based on several publications, but in particular this one put together by the team in Pittsburgh where they essentially explored the outcome of patients with sepsis before and after the implementation of a sepsis improvement program. And the sepsis improvement program in question is the Rory regulation as it was implemented in New York state following the death of a 12-year-old boy. And so the team led by Jeremy Kahn from Pittsburgh explored whether this mandatory implementation had made a difference. They compared the outcome of patients with sepsis before and after this implementation. And as you can see here in New York state, clearly there was an improvement in outcomes. 30-day mortality improved, was lower after the implementation. The risk of ICU admission was lower, and the patients stayed in hospital for shorter. Now you may say, well, this could have happened anyway because people were more aware of sepsis and there was much better recognition throughout the world. So they compared this with control states where the Rory regulation had not been mandated. And as you can see here, indeed, across the world and certainly across these states, there had been some general improvement, but overall it was felt that the improvement in New York state was better and they felt it was probably related to this mandatory implementation of a new program. And this is clearly good news. This plus some other reports convinced the surviving sepsis panel to issue the recommendation as you saw it. Now the second new recommendation was a recommendation against the Q-SOFA score as a screening tool. As you all know, the Q-SOFA score is a three-point scale which was put together by the sepsis three panel when they put the sepsis three definition together. And this consists of these three parameters. And it was really developed to predict mortality in patients with sepsis. But it happens that across the world it has been used as a screening tool to identify patients with sepsis. In fairness, it was never intended for this purpose. But people have used it and certainly reports are in the literature which prompted Professor Povoa from Portugal and others to look at the existing publications and look at the performance of the Q-SOFA as a screening tool. So they looked at all articles comparing the Q-SOFA with SIRS in diagnosing sepsis. And here are the 10 articles. They looked at patients. These articles cover patients in the ED, on general wards, but also patients in the intensive care unit. And to cut it short, if you look at sensitivity, the majority of the studies favored SIRS. And when you look at specificity, sadly there was only one study. And so not surprisingly, the team concluded that SIRS was a much better tool compared to Q-SOFA. And several other investigators looked at the same topic and compared Q-SOFA with not only SIRS, but also NEWS and NEWS and came to similar conclusions. Here is one of the landmark papers conducted by the team in Chicago, where they again applied the different scores to patients in the emergency department and on the ward. And again, the Q-SOFA score performed less well compared to NEWS and NEWS score. The team appropriately concluded that traditional warning scores were much better and more accurate compared to Q-SOFA at identifying patients. These two studies, together with a few others, underpin the recommendation from the surviving sepsis panel. So it all seems to make sense. The only trouble or challenge in real life at the bedside is obviously that none of these tools really help us identifying patients with sepsis. They were all warning tools. They were tools to identify the deteriorating patient, the patient who is likely to progress. But they were never developed and put together as tools to really identify patients who are developing sepsis. Ideally, that's what we want, we would like a tool that tells us that the patient is developing early organ dysfunction and has an infection. These tools, at the moment, do not exist. And the tools which we use are not really sepsis-specific. They may raise our suspicion, but they can't diagnose sepsis. And so that's the dilemma. Although the surviving sepsis guideline urges us to use tools to detect patients urgently in real life, we actually don't have any appropriate tools. Now obviously, there are automated systems, and some get promoted. And certainly, the EPIC system claims to have a sepsis recognition system. It has been implemented in many hospitals across the US, but not only in the US, but also across the world. And certainly, in my hospital, we're in the middle of implementing EPIC, including this warning system. Now when you look at the performance reports for this tool, unfortunately, they don't exist. And this prompted a team here to do their own performance evaluation. So they evaluated the score in patients admitted to the hospital during a 12-month period. Sepsis, similar to the EPIC system, was defined by these traditional criteria as listed here. And in their 12-month evaluation, they looked at the data and the outcome of more than 27,000 patients who had one or more hospitalization. This was a group of patients where the majority were women, had an average age of 56, presenting to the emergency department. And in this cohort, during this 12-month period, indeed, 7% of patients developed sepsis. When they looked at the EMS score, then as you can see here, depending on where you set your cut, if you have a lower cutoff, a lower threshold, then you may have higher sensitivity, but you have a very low specificity. This is a problem with any score, but the EPIC system recommends to use a score of around 6. So that's what the team did. Using a score, a cutoff of 6, meant that, indeed, of the 2,550 patients with sepsis, there were 7%, 187 patients, who had a score of 6 or higher. And at that time, they were not on antibiotics, antibiotics had not been commenced, and they subsequently developed sepsis. So in these 7% of patients, a score could have helped. It would have alerted the clinicians. There were also a large number of patients who had sepsis, and the score was never 6 or more. And they were not identified by this system. And furthermore, there were a large number of patients who had a score of 6 or higher, but never developed sepsis. And clearly, they would have been alerted by the clinician, and they would have been evaluated. And the clinicians would have spent a lot of time identifying or investigating them, and then finding they didn't have sepsis. And so this team felt that this particular score, which is being rolled out across the world, actually had poor discrimination and calibration, and led to unnecessary workload, and probably promoted alert fatigue. So this conclusion is not particularly promising, given that, across the world, we are rolling out this score, including in my hospital at the moment. Is there anything else on offer? Well, the Cochrane team also looked at automated systems, and they looked at existing reports up to 2017. Sadly, they could only find three randomized controlled trials, of which two were only available in abstract form. So the data are very, very limited. And in this limited data, there wasn't really any convincing message or convincing data to support automated systems. And so all they could conclude was it's unclear whether automated systems help or not, because we don't have data. And it's also unclear whether any of these systems help clinicians at the bedside. And that's coming to the end of where we are. Essentially, we have great recommendations, which I fully support. But in real life, we don't have a sepsis screening tool, although the Surviving Sepsis Guideline recommends us to use one. And we also, in fairness, struggle, really, to, if we were asked to find and put a tool together, it would be difficult, because we don't have the right diagnostics and standards for diagnosing sepsis and organ dysfunction. And what we clearly need to do is invest more in the development of the right tools that are accurate, worker, clinician-friendly, and identify the right patient without causing unnecessary work and alert fatigue. And with that, I want to end. And thank you for your attention. And without any further delay, I introduce my friend, Sheila Mitra from Mumbai. She's the president of the Indian Society of Critical Care Medicine. And she's an expert in fluid management, is going to tell us about individualizing fluid management. Sheila, the floor is yours. Thank you, Marlies. Good afternoon, friends. And thank you for participating in this session and my talk on individualizing fluid resuscitation. I'm Sheila Mitra, professor of critical care, working at the Tata Memorial Hospital in Mumbai, India. I plan to cover the evidence gaps and challenges in fluid resuscitation in patients with sepsis and septic shock, and also talk about how you could assess fluid resuscitation, what are the limitations, and how you could individualize fluid resuscitation. Now, volume expansion is usually the first line of therapy when you have a patient who presents you with septic shock. And when we give fluids to the patient, and you've seen the Frank-Starling curve several times, we presume that the patient is on the steep part of the Frank-Starling curve, which means by giving a certain amount of fluid, you're going to have a proportionate increase in the stroke volume. However, the situation may not be the same after you've given a significant amount of fluid. And you don't have one Frank-Starling curve, but a whole family of curves, depending upon how your cardiac function is. So in some patients, you might give the same volume of fluid, but you may not have a proportionate increase in the fluid. And this is what we mean, look at, when we assess fluid responsiveness. So when we have a patient presenting to us with septic shock, really determining which patient is going to be fluid responsive is really a challenge in the intensive care unit. Because on one side, you don't want to give too much fluid and push the patient into pulmonary edema. Whereas on the other side, you do want to restore organ perfusion and correct the hypotension. So when you ask the question that, should I be giving more fluid, what you really want to know is that, will the cardiac output increase after fluid loading? Because if the cardiac output doesn't increase, there's really no benefit of giving any fluid, and you might actually produce some harm to these patients. So when you look at fluid responsiveness, what we mean is that this is a state when administration of fluid will lead to improvement in the stroke volume, and hence the cardiac output of the patient. And Maliz has just talked to you about the most recent iteration of the Cardiac Surviving Sepsis Guideline, which was published in 2021. And I'm going to talk about the fluid resuscitation recommendations. And you're all familiar with this. And it talks about the treatment and resuscitation should begin immediately. And then this recommendation of giving at least 30 mils per kg of intravenous crystalloid fluid within the first three hours of resuscitation. And this is one of the most controversial recommendations and the most debated recommendations in the Surviving Sepsis Guidelines. And of course, using dynamic measures, which I'll talk about in a bit of time. So this recommendation of 30 mils per kg of crystalloid is actually based on one retrospective analysis of adults presented to the emergency department with sepsis and septic shock, in which it was shown that failure to receive 30 mils of crystalloid within three hours of sepsis onset was associated with an increased odds ratio of in-hospital mortality, delayed resolution of hypotension, and increased length of stay, irrespective of the comorbidity. So this is one retrospective study. In fact, from the previous iteration of 2018, they've actually down-regulated. And they've made it, you know, because the evidence is low from we recommend to we suggest now for this 30 mils. So when you're giving 30 mils per kg to a young, healthy person, I mean, a young boy with septic shock with no comorbidities, it's not the same as giving it to an elderly person who has comorbidities, who has, say, renal failure, and also has cardiac failure. So you can't really have a one-size-fits-all approach to every patient who presents to you with septic shock. And I was fortunate to be part of this group, which wrote this editorial soon after the Surviving Sepsis Guidelines was published about how we could equilibrate, about equilibrating the Surviving Sepsis Campaign Guidelines with individualized care, how you could work within the same framework, and you could actually individualize your care for the patient. And I'm, you know, we came up with about 20 recommendations, but I'm not going to be talking about that. I'll allude to the one on fluid administration. And we recommend individualizing initial fluid resuscitation, because no single formula can be applied to all patients, and fluid requirements vary substantially, again, depending upon what is the source of sepsis, and also what is the preexisting cardiovascular function. So you can't, you know, 30 mLs may be too little for someone, and it may be too much for another individual. So this is a very nice, elegant paper that comes from Professor Tabool about how you could, you know, fluid resuscitation during early sepsis, and how you could individualize care within the framework of the Surviving Sepsis Guidelines. So you're talking about 30 mLs given over three hours. So three hours is a very long period of time. Of course, we're going to reassess our patients in between. So you, you know, when you have a patient presenting with septic shock, you could infuse about 10 mL per crystalloid to start with, and then assess the patient. You don't have to give 30 mLs and wait for three hours. So after about an hour, you can see whether there is any worsening of the tachypnea or fall in oxygenation, and perhaps you might consider not giving more fluid. Or you can have a situation where you have, there's a patient has evident, you know, fluid losses, abdominal, of abdominal origin, fever, then, you know, clinical signs of sepsis, I mean, tissue hyperperfusion still persist. So you may consider giving the remaining amount of fluid. And of course, if the shock persists, then you can consider assessing fluid responsiveness. So this is a more individualized way by which you can still fulfill the 30 mLs, but you're assessing in between and deciding whether it's, you know, being more cautious before giving further fluid management. Then coming to the next recommendation about using dynamic measures of fluid, fluid to guide fluid resuscitation. And the tests that they've talked about are pulse pressure variation, stroke volume variation, using passive lead raising, or even giving a fluid bolus and looking at the increase in stroke volume. And of course, echocardiographic measurements. I'll talk about a few of these tests and what are the limitations of this test. I'm sure you've seen these kind of dynamic changes in the art when you put in an arterial waveform during mechanical ventilations, during inspiration and expiration. Now, this is normal physiology. But of course, you see big swings in the arterial waveform when a patient is fluid responsive. And this is the principle that's used when pulse pressure variation and stroke volume variation is calculated. You take the maximum minus the minimum, and you divide it by the mean, and you get this kind of number on the monitor, a cutoff value above which the patient is fluid responsive. And if it's not, with a reasonably good specificity and sensitivity. So pulse pressure variation and stroke volume variation are one of the most widely studied tests in clinical practice. And if you look at the three meta-analysis that have looked at these tests, you have a reasonably very good area under the curve. However, these tests are not perfect. There are limitations with the use of pulse pressure variation and stroke volume variation. A lot of condition in which you get false positive and false negative readings. And one of the most common in clinical practice is patients ventilated with low tidal volume. And today, we're ventilating patients not only with ALI and ARDS, but also patients with sepsis and trauma with using much gentler on the lungs, even in the operating room. So we're using about 6 mL per kg tidal volume or even lower. Now, these tests are excellent. They work very well at 8 mL per kg. But when the tidal volume is less than 8 mL per kg, that amount of tidal volume is not able, is inadequate to elicit the changes that are seen with these tests to interpret the results. And this is the main limitation. So you have an excellent test, but it doesn't work at tidal volume less than 8 mL per kg. And in our unit, this is what we tried to do. We developed a new test called the tidal volume challenge. And what we hypothesized is that you have an excellent test, pulse pressure variation, stroke volume variation. Works very well at 8 mL per kg, but doesn't work at 6 mL per kg. So why are we throwing the baby with the bathwater? Why don't we try to overcome this limitation with the low tidal volume? And that's what we did when we developed this test called the tidal volume challenge. And what we did is we said, OK, it works at 8. It doesn't work at 6. So when you're ventilating patients at 6 mL per kg, why don't we transiently increase the tidal volume from 6 mL per kg to 8 mL per kg, and then look at the pulse pressure variation or the stroke volume variation? That was our hypothesis. As you can see, the pulse pressure variation and stroke volume variation at 6 mL per kg, the area under the curve was very poor. At 8 mL per kg, as we hypothesized, it was great. But even better than 6 at 8 mL per kg was the delta. That is the difference between the value at 6 mL per kg and 8 mL per kg, which was a better predictor of fluid responsiveness. And the cutoff value was 3.5 for pulse pressure variation and 2.5 for stroke volume variation. And I'll just tell you how simple it is to do this test without using a continuous cardiac output monitor. You can get pulse pressure variation on the monitor if you have a simple arterial line, which you would have in a patient presenting with septic shock. So that's the monitor. And you can see the pulse pressure variation. The value is 8. And this is your ventilator, patients being ventilated with a tidal volume of 6 mL per kg. 270 is the tidal volume. 45 is the ideal body weight. And now we've increased the tidal volume from 6 mL per kg to 8 mL per kg, which is 360 mLs. And you will have to look at the pulse pressure variation. We do this for about one minute. And you can see that the pulse pressure variation has increased from 8. It has become 10. And not more than one minute. And from 8 to 10 to now it is 13. And after one minute, of course, you have to bring the tidal volume back to 6 mL per kg. So you can see the pulse pressure variation increased from 8 to 15. And after one minute of increasing this tidal volume, we bring the tidal volume back to 6 mL per kg. And that's what we're doing. And it's very interesting. Because when you bring the tidal volume back down, the pulse pressure variation also goes down. Very quickly, how to interpret this test, you started with 8. And after one minute of increasing the tidal volume to 8 mL per kg, the pulse pressure variation became 15. So 15 minus 8 is 7. This is more than the cutoff value of 3.5 for pulse pressure variation. You can use this with a good specificity and a good sensitivity. And it doesn't need any continuous cardiac output monitoring. And that was a proof of concept study that we had done in our unit when we developed the test. But subsequently, you have several studies now, more than 20 studies, and a recent meta-analysis of these studies, which has shown that the tidal volume, pulse pressure variation, can be used with the tidal volume challenge. And this can be used not only in patients with low tidal volume, but also with low compliance. And also when there is moderate PEEP. But still needs to be tested in patients with spontaneous breathing, and in patients with, you know, who are in prone position. Echocardiographic variables, of course, we can use SVC variability, IVC variability. We can look at the aortic rose velocity. But the most popular and used at the bedside is the trans-thoracic echocardiography, where you look at respiratory variations in IVC diameter. And then, of course, depending upon which formula you're using, there are different cutoff values. Not very good during spontaneous breathing, more reliable during controlled mechanical ventilation. Now, this is one of the largest study that's compared all these echocardiographic indices to predict fluid responsiveness in ventilated patients. It was a multi-center study that was conducted in France, which compared all these indices. And unfortunately, the results were not very good. And you can see that the area under the curve, the highest was about 0.74. And that was for the delta SVC variability. So ultrasound sonographic tests are very easy to do at the bedside, non-invasive. You don't need to put in any catheters. The learning curve is also not very steep. After doing a couple, you can start looking at IVC variability. However, you must remember that the reliability of these tests are not very good. And it's very interesting to note that two-third of these patients were actually ventilated using low tidal volume. So the low tidal volume, which affects most of the tests that use heart-lung interactions, also affects those patients who also affects the results of the IVC variability. And then, of course, we have the passive leg-raising test. This is from the group of Professor Tabool. And the beauty of this test is it doesn't depend on heart-lung interaction. So it can overcome most of the limitations, like low tidal volume, spontaneous breathing, arrhythmias, et cetera. But you have to do it very properly. You have to start from semi-recumbent position. And you have to move the bed and not the patient, because you don't want any sympathetic stimulation. The only disadvantage is that you need real-time cardiac output to monitor this. You need an increase of cardiac output more than 10%. The test is very easy to do. But for the interpretation, you need cardiac output monitoring. And they have the area under the curve, and this is a meta-analysis that was done by the same group. And a good threshold, easy threshold to remember, was 10% increase in the cardiac output after performing the passive leg-raising with a good specificity and sensitivity. And this is a recent meta-analysis which looked at all these recent tests that we're using for predicting fluid responsiveness. And if you see, most of the commonly used tests, pulse pressure variation, stroke volume variation, even tidal volume challenge, passive leg-raising have a good area under the curve and seem to be reasonably reliable. But what most people say is, OK, you have these tests. But what about outcomes? Do you have any outcome data? And this is a very nice study from Argentina, and this looks at patients in a multi-center prospective cohort study that looked at fluid responsiveness and outcomes in patients with sepsis. And if you look at the independent determinants of mortality according to logistic regression, use of dynamic tests of fluid responsiveness was significant. So you actually have outcome data now for using tests with fluid responsiveness. So when you ask yourself, when should I give fluid? It's not just because the patient is fluid responsive. That's only one part of the question. Being fluid responsive doesn't mean that you need to give fluid to the patient. So you and I may also be fluid responsive at this time of the day, but we are not in acute circulatory failure. So first, you should have the presence of acute circulatory failure. Then, of course, you need to test whether the patient is fluid responsive or not. And even if the patient is fluid responsive, if there are any risks, there should be no major risk of giving fluid to this patient. For example, if he has a very bad heart or he has very severe ARDS, the patient may be fluid responsive, but I may still not consider giving fluid and perhaps give vasopressors to this patient. So I'll end with my take-home message being that fluid responsive therapy is usually the first line of resuscitation in patients with septic shock, unless there's some contraindication or there's profound, a patient is in very severe, a life-threatening hypotension, or you have a very low diastolic pressure, a very vasodilated patient, in which you may consider giving vasopressors up front. And Professor Bauer is, of course, going to be talking about that. And the SSC guidelines have suggested now that you use 30 mils of crystalloids over three hours in these patients. But remember, this is low quality of evidence based on one retrospective trial study. So an individual's approach may be followed, where you could give an initial amount of fluid to this patient, assess fluid responsiveness, and then consider whether you want to give more fluids to these patients. And you could use the following dynamic tests, but remember that they do have certain limitations. And most importantly, even if this patient is fluid responsive, remember, don't give fluid if there is an additional risk of giving fluid to this patient. Thank you very much for your attention. Thank you. And without any further delay, now we move on straight to a vasopressor talk given by Professor Bauer, pharmacist from the Cleveland Clinic and an expert in cardiovascular support. And clearly these talks complement each other very well. So, Professor Bauer, the floor is yours. Thank you very much. I get the opportunity to follow these two fantastic talks and I'll try to keep us on time here. Get the opportunity to talk about vasopressors, which naturally follow after fluids. I'm a critical care pharmacist from Cleveland Clinic and will disclose that my institution has received funding from NIGMS. The objectives here are to talk about the gaps in the evidence for vasopressor management and hopefully leave you with some idea of how to overcome these gaps in your practice at the end. I had intended to use the audience response system, but we'll omit doing so just for the sake of time. When thinking about all of the questions that we could talk about for vasoactive agents and what the gaps are, this is just some of the list that I came up with. But of course, you don't wanna hear me talk about all of these things. You wanna enjoy this fantastic location. And so I'll eliminate some of them off the bat because they don't have to do with vasopressor management and progressive shock. And I also know that there's a session tomorrow that deals with refractory septic shock and initiation of vasopressin will be discussed in angiotensin II. So if you wanted to hear about those things, please join us tomorrow. So that leaves me with some interesting questions to discuss and that would be vasopressin deescalation or vasopressin cessation and the role of epinephrine. We know that the surviving sepsis campaign guidelines recommend or suggest the use of vasopressin as an adjunct to norepinephrine in patients in whom the MAP cannot be achieved with norepinephrine alone. However, the guidelines make no mention of how to cease vasopressin in patients in the recovery or deescalation phase. There's only one trial that has evaluated this question. That trial is called the DOS trial. It was a relatively small trial enrolling only 78 patients. In this trial, they randomized patients to have norepinephrine ceased first or vasopressin ceased first. If norepinephrine was stopped first, the odds of subsequent hypotension were sevenfold higher. However, in an individual patient data meta-analysis that included five observational studies and this randomized trial, if you stopped norepinephrine first, the odds of hypotension were lower which creates an interesting juxtapose between these two findings, between observational data and a trial. However, if vasopressin was stopped first, the total duration of vasopressin was less. ICU length of stay was no different and no difference was detected in short-term mortality. So this offers an interesting question of whether we should be stopping vasopressin first or norepinephrine first. There's a disconnect between observational studies and a randomized trial. There's also a disconnect between a surrogate measure that we think matters like hypotension and ultimate patient outcomes like length of stay and mortality. And the question arises whether hypotension after stopping one of these agents actually matters for patient outcomes. A related but different question is how should we stop vasopressin? In observational studies, vasopressin is typically used at one single dose and turned off when we make the decision to stop vasopressin. No randomized trial has asked this question but two observational studies have evaluated it. One of the observational studies was on the larger side and included about 1,300 patients. The other was on the smaller side. The two methods that were evaluated was that abrupt stopping of vasopressin or weaning it off or titrating it off like we would with say norepinephrine. In these two studies, no difference was detected in the risk of hypotension if we abruptly stopped vasopressin or titrated it off but the duration of vasopressin was lower if it was abruptly stopped. Additionally, no difference was detected in ICU length of stay if we abruptly stopped or weaned off vasopressin. So after all of this mention of no difference detected, you may be saying to yourself, it doesn't matter, I can just do what I want, that makes it so easy. However, there are some important nuances here that are worthy of consideration. The first of which is in the randomized trial, the investigators evaluated copeptin. Copeptin is a stable fragment on the C terminal of provasopressin and its concentration mirrors that of vasopressin concentrations. The advantage of copeptin measurement is it's much more stable than vasopressin and easier to measure. These investigators found that in the patients randomized to have norepinephrine stopped first, there was no difference in copeptin. I'm sorry, I'm not even touching it and it's advancing. Apparently, they want me to move on. In the patients that had norepinephrine stopped first, there was no difference in copeptin levels in those who developed hypotension or no hypotension. However, my goodness, however, in those that had vasopressin stopped first, the copeptin levels were lower, suggesting that these patients did not have restoration of their endogenous vasopressin access and there could be some role of using copeptin to help us make that decision. Vasopressin and copeptin levels aren't readily available at the bedside, but what is readily available is the knowledge of knowing if norepinephrine was stopped first or if vasopressin was stopped first, meaning if you're going to stop vasopressin, you know if norepinephrine is still running or if you only have vasopressin. In a subgroup analysis of the larger observational study that evaluated the cessation question, there appeared to be a subgroup effect based on the discontinuation order for whether vasopressin would be abruptly stopped or weaned. In patients that had norepinephrine stopped first, those that had vasopressin abruptly stopped seemed to have a shorter ICU length of stay, but the difference was reversed in patients that had vasopressin stopped first and those that had vasopressin weaned off seemed to have a shorter length of stay. Now again, this is hypothesis generating, but it could be applicable to practice. I'll move on to discussing the role of epinephrine. Epinephrine has been in use in clinical practice for over 100 years, but relatively few patients have actually been enrolled in trials that have evaluated epinephrine use. These are data from a relatively recent meta-analysis that only included about 1,200 patients of any shock type that received epinephrine versus a control, and in this study, they did not detect a difference in mortality at the longest follow-up time or at 28 days or 30 days, and when they used sepsis subgroups of sepsis or non-sepsis, the point estimate for sepsis favored epinephrine, but confidence interval overlapped the null, and for non-sepsis, the point estimate favored the control. The authors concluded here that the data support no increased risk of mortality with the use of epinephrine in circulatory shock. I think that that conclusion can be supported, but maybe a better conclusion here is that the evidence is uncertain because the confidence intervals are wide and include clinically important benefit and clinically important harm for these patients. The two studies that drove this decision are probably well-known to you, one called the CAT study, the other, the CATS study. The CAT study was a study evaluating patients with all shock types, about half of which had septic shock. The primary outcome was the time to achieving the MAP goal. The study didn't detect a difference in this primary outcome, but they did detect a difference in adverse effects with epinephrine. Specifically, patients were withdrawn from the study more frequently with epinephrine because of tachycardia and hyperlactatemia, and this is reflected in their measurements of lactate and heart rate for these patients. In a trial evaluating patients with septic shock called the CATS study, patients were randomized to epinephrine or the combination of norepinephrine with dobutamine. Primary outcome was 28-day mortality here, and they didn't detect a difference in 28-day mortality. They did, again, see higher lactate concentrations with epinephrine, but it was transient and only on the first day, but arterial pH was lower with epinephrine. One could conclude from this trial there's no difference in mortality, but I think there's some important nuances with this trial that deserve mentioning. The first is that all patients were on norepinephrine at baseline, so this trial well aligns with adjunctive use of epinephrine and not first-line use of epinephrine. Additionally, in this study, and I didn't know this from reading the manuscript by itself, all patients had the mandatory inclusion of dobutamine at baseline, and the dobutamine could be decreased or titrated off if the cardiac index was adequate. So we don't know if the use of epinephrine versus norepinephrine alone is supported by this trial. And lastly, the trial was powered to detect a 20% difference in mortality, and that's really the only conclusion we can make is that there is no more than a 20% difference in mortality between these approaches, and that probably exceeds what we would agree is a clinically important difference, and that's reflected in these wide confidence intervals here for the mortality difference. I think it's important to also note that observational studies have found the opposite effect with epinephrine. In one study of patients with septic shock, the model suggested that epinephrine was associated with higher risk for mortality, and also in cardiogenic shock, there are models that suggest epinephrine is associated with higher risk for mortality. You may be asking why I'm talking about cardiogenic shock here in a sepsis talk, but that's really why we use epinephrine, right? We have an assessment that a patient has low cardiac output. We wanna give epinephrine to raise the cardiac output, and that is akin to how we think about epinephrine for cardiogenic shock. One relatively recent trial evaluated this question in cardiogenic shock. This study took patients who were in acute cardiogenic shock from acute MI after successful reperfusion and randomized them to norepinephrine or epinephrine, and the study was stopped early for safety concerns. We see here that cardiac index transiently increased with epinephrine, but only over the first few hours, and not over the entire study timeframe, but mortality was higher with the use of epinephrine, and this was driven by a higher risk for refractory shock with epinephrine. The rationale here is hypothesized to be with increased myocardial oxygen consumption because the cardiac double product was higher with epinephrine. The double product is heart rate times stroke volume, so even in a reperfused myocardium, you can still see potential ongoing ischemia with the use of epinephrine. Bringing this back to the sepsis literature, I think there's an instructive animal model that can help us here. This is an animal model of a rat model of cecal ligation and puncture where they induce sepsis in these animals. They did measurements of hemodynamics, instrumented them, fluid resuscitated them, then randomized them to receive norepinephrine or epinephrine as the vasopressor titrated to the same goal. Because the absolute values don't apply to humans, I'll just show you the relative changes from baseline. The animals randomized to norepinephrine versus epinephrine did not have a difference in cardiac output, which doesn't make sense based on what we know from the pharmacology. And you may be saying, oh, cardiac output is afterload dependent, but they actually also evaluated preload recruitable stroke work, which is a load-independent marker of myocardial contractility, and there was no difference between these two agents. Heart rate was higher though with epinephrine, so maybe cardiac output doesn't increase because there isn't adequate diastolic filling time to see those inotropic effects of epinephrine. Also, the pressure volume area was higher with epinephrine, which is also a marker of myocardial oxygen consumption. So when we're thinking about using epinephrine in this sepsis myocardial dysfunction model, we might wanna second guess if it's actually going to achieve the goals that we are seeking. I'll wrap up with a few points for practice. I think we should consider stopping vasopressin before stopping norepinephrine because here in the U.S., or at least mainland U.S., the cost of vasopressin is quite high, and so stopping it earlier can help us save some drug costs. This can lead to shorter vasopressin duration, maybe at a risk for some hypotension, but the effect of that hypotension is unclear. Generally, that just means we increase the norepinephrine dosage. And if we choose this pathway of stopping vasopressin first, then we probably should wean vasopressin off as opposed to abruptly stopping it. I'll submit that the role of epinephrine in shock is quite unclear, that the outcome evidence is uncertain, and the pharmacology suggests we should use it as an inotrope, but the data don't necessarily support that it achieves the intended effect, and trials are needed for this drug. Thank you. We will now move to the final talk of the session, and it's an honor to introduce Professor Vincent Liu, a research scientist from Kaiser Permanente in California, also an expert in AI. But now the talk will focus on personalized medicine in sepsis patients. Vincent, the- So I have the opportunity to follow these wonderful talks and step away from the surviving sepsis guidelines. I was quickly checking my phone to see whether there was a section on this, but I don't think so. So I feel okay about that. So this is more high-level considerations about, you know, what are we thinking in personalized and precision medicine. And I will start with the disclosure that I'm not a translational scientist. So I'm taking the perspective of someone who's deep in DEHR data and observational data, and from a health system perspective, rather than with deep knowledge about transcriptomics and other methods like that. These are some of my funding sources. And I think that this is one of the most informative images that I have when it comes to heterogeneity. And so this was the cover art for the issue when the sepsis 3 definitions were published in 2016. I apologize. It's too small to see from the back, but I think you get the picture that this is what we would like to be seeing on our monitors. Organ dysfunction, heat maps, and cytokine profiles, laboratory data, vital signs, temporal trajectories of illness, inflammatory immune status, host-pathogen interactions, population profiles, and comorbidities. And I think this is an aspiration of where we would like to be. So that's gap number one. We don't have a monitor, nor do we actually even have most of these things well-defined. Number two is, even if I made such a monitor appear in front of you, would any of us know what to do with this? And the answer today is no. But it's still an aspiration for us to get to this point. All of what you've heard are kind of recommendations. But when we practice, we understand that, as Sheila said, the risks need to be balanced against the evidence for my individualized patient. And so certainly only a protocolized approach, without considerations for the heterogeneity present within our individualized patients, is going to lead to harm for our patients and not benefit. But we poorly observe heterogeneity. And we've known this for a long time. This editorial, Negative Trials in Critical Care, why most research is probably wrong, gets into why we don't understand the underlying physiology and biology that's actually happening in sepsis. And that hampers our ability to identify efficacious interventions when it comes to clinical trials. John Marshall from Toronto had a similar piece several years ago, where he looked at many of the trials which have examined immunomodulatory therapies over several decades now and examined, again, or described that we don't have a good understanding of what's happening in the immune system. And that hampers our ability to treat patients. And then going back to 1996 and before that, Roger Bone, why sepsis trials fail. And again, just again, harping on this concept of heterogeneity, a lack of understanding. And we've thought about that as kind of this tension over this past decade, which is this tension between lumping and splitting. So lumping would be to say, we recommend with weak, low-quality evidence that we should give everyone 30 cc's per kg. And splitting would be to say, well, that's actually not a great approach because we need to individualize fluid therapies. And I think that has to do with the fact that we have some pros to lumping, which is that we give a certain standard that people can match. We get to produce some set of quality measures to drive towards. But along comes with it the downsides, which is that in some number of patients, that's almost certainly going to be too much fluid. I think we are moving from a place where there was advantages, which is that every doctor can treat a patient however they feel like, so ultimately individualized. But the downside of that was that we knew and have known that a large number of patients were under-resuscitated significantly. And that's what we've seen as one of the dominant kind of changes in the way we approach population care and sepsis over these past 20 years. This is reflected in this really nice piece from David Maslove and a whole number of authors in critical care where they seek to redefine critical illness. And this speaks to the phases in terms of definitions, and then kind of what are the advances technologically and pharmaceutically, and then with the severity scores that we've seen over these past decades. And ultimately, lumping was a strategy for this acceleration phase, where we're trying to define things, where we're trying to seek consensus, so that we can bring everybody in the field up to a certain level of practice. But it's unlikely to serve us as we move forward into the age of precision, which is where we need to begin to understand the unique characteristics and combination of characteristics within patients to treat them well. This is also nicely laid out by Pratik Sinha, who's now at St. Louis, but the challenge has been that traditionally here we are conducting randomized control trials or observational studies in sepsis and ARDS, and we are homogeneously applying an intervention to a very heterogeneous target. What we really need is some aspect of biologic phenotyping, and that could be from a variety of different biologic processes or molecular bases. We need them to be clinically implementable, so at the bedside we're able to act on them. We need to test for this heterogeneity, so-called heterogeneity of treatment effect, the fact that even that on average an intervention, for example, is shown to have no effect. There are some patients who benefited tremendously from it and others who were harmed by it. And then we are ultimately going to need to turn that into bedside tools, both new therapeutic targets as well as randomized trials that are leveraging functional assays. And in the paper by Maslov, David, he goes through a kind of schema for this, which is that in critical care, that sepsis and ARDS and other types of critical illness are actually a collection, they're syndromes, and they're likely to be defined by some kind of underlying biological characteristics. What we need to do is unpack those. We need to use tools like machine learning in order to identify specific physiologic states of interest. And then ultimately he calls these treatable traits. And what we need to do is understand what the treatable traits in our patients are rather than kind of a blanket approach to treating sepsis, treating ARDS, and other conditions. Here you can see there's pancreatitis, trauma, surgery, infection. And of course, we need assays. We need point-of-care tests. Even if we were to actually have that monitor available, we need to then be able to act on it, know the thresholds at which we would act, and then test those within novel trial designs. And it's not that we currently lack axes of heterogeneity. And so Pratik nicely goes through a bunch of these. And so for those of you who are familiar with some of these phenotyping or endotyping or subgroup work that's in the literature, you'll be familiar with some of these. But call out late Hector Wong's work in pediatric septic shock and examining subclasses of sepsis that appear to be more responsive or harmed by corticosteroids. And that's been a theme across many of these, that many of them are examining axes of immune dysregulation and underlying inflammatory states of patients. Again, lots of examples when it comes to immunophenotyping and then protein biomarkers as well. So many of these are in the pipeline. They've been examined in translational data sets. And then the kind of unsupervised or empirically generated phenotypes have then been tested in existing randomized trial data. And so one example, I'll just dig into a little bit more. And again, I apologize. This is small. But this is from Chris Seymour's group out of Pittsburgh. And this was a pretty high profile paper in which they used both observational data as well as several randomized trial data. They use clinical data and biomarkers in order to come up with putative subtypes of sepsis. And so here they have a so-called Seneca data set. And then they had data sets from several randomized trials, which they used to then examine what the effects of these putative potential subgroups would be. And they uncovered so-called alpha, beta, gamma, and delta subtypes, which varied substantially in terms of their need for mechanical ventilation, the rate at which they were admitted to the ICU, and the in-hospital mortality across these groups ranged from 2% in the alpha group all the way up to 32% in the delta group. They then examined what these subtypes, what the mortality would have been in, for example, the PROWESS trial and the PROCESS study, which were both trials examining EGDT, and showed that they did very clearly differentiate the likelihood of mortality among patients who were randomized in those treatments. And then finally, they said, we took all of these variables and used an unsupervised method called latent class analysis to show which of these factors were more well-represented, for example, in the alpha subset compared to the delta subset. And so again, there is much work like this underway. Many subgroups have been proposed. We have not yet seen the one subgroup or framework to rule them all. But clinicians like me and many of us are hoping for that, rather than getting into all of the nitty-gritty of transcriptomics, genomics, and protein biomarkers. And I think Pratik put it well. I mean, what we lack is kind of the clear impetus to turn these into bedside tools, partly because they're currently experimental. We lack the prospective data we need. Again, even if that monitor were available to us, we would lack the knowledge about how precisely to treat them. But we suspect that clinical trials are emerging today, and many of them may actually produce some insight into when to use corticosteroids, for example, as opposed to the evidence that we have today. But then there's other considerations, feasibility. Are we going to have these biomarkers? Are we going to be able to overcome clinician skepticism? Are the subtypes that have been proposed, are they even sufficiently complex? And do they actually reflect the complexity of underlying pathophysiology? And then ultimately, there's a lot of challenges, because often these are measured at a single time point. And so we don't understand the trajectory of illness, nor do we understand when all patients show up in the ED, some of them are on this trajectory, others are on this trajectory, others are stable. So we lack a good temporal phenotyping for patients, which is going to be a challenge as well. But my hope is that if I'm invited back in five years from now, that someone will be selling this monitor, and we'll be able to chunk out pieces of it and talk about surviving sepsis guidelines that target each of them. Thank you. Thank you. Thank you for giving us some hope for the future.

sepsis

screening

identification

fluid resuscitation

individualized management

vasopressors

epinephrine

personalized medicine

improving outcomes