Okay, thank you very much for joining us this evening. I know that most of you had a big hike to get here. It's a very large, expansive conference center. But thank you, and I hope you're enjoying your tapas and your wine. It's my very great pleasure, and I sincerely mean that, to introduce you to our first speaker, Dr. Stephen LaRosa. He has, and I'm sure he'll express it and you'll recognize, a true passion for sepsis and antimicrobials, and I think you're going to enjoy very much what he has to present to us this evening. So he's currently at Beverly Hospital, part of Leahy Health Beth Israel. And he's got a, as you can see, he's got a glowing resume. So I'm not going to waste any more time, and I'm going to hand over to Dr. LaRosa. Thanks, Jane. I recognize some friendly faces from the past, so it's nice to see that you've come out for this. Now I'd like this very much to be kind of a walk down what's happened throughout history in sepsis and how it brings us to the current technology that we'll be talking about tonight. It's really about the challenges of making the diagnosis of sepsis and how we can harness a new way, a new technology to do so. So a lot of you in the room are probably old enough to remember the original sepsis criteria. I've affectionately called it the Las Vegas criteria because the interesting story I've heard from some of my teachers and mentors were, as a useful construct for clinical trials, it was a group of sepsis experts sitting in a room arguing about what sepsis was, and some people would fight and go out of the room and come back, and finally they said, let's just write something down and we'll come back and we'll revisit it in a few years. And for many years, this was the definition. It was the presence of an infection in concert with at least two SIRS criteria, sepsis inflammatory response criteria, which was temperature, heart rate, respiratory rate, and white blood cell count aberration. Having at least two of those gave you sepsis, and if you had an organ failure going along with that, including shock or the need for mechanical ventilation, that constituted severe sepsis. And this is really a useful Venn diagram depicting what the problem is, right? So infection and sepsis are not the same thing. There's an overlap, and there's a large overlap with SIRS, and the problem is there are many noninfectious entities that cause SIRS. On teaching rounds, we'd often make the joke that when I ran up three flights of stairs, I had SIRS. I had an elevated heart rate and an elevated respiratory rate. Not that I knew of. So here's a list of just some, and this is not an all-encompassing list. There's tissue injury, hematomas, MIs, PEs, transplant rejection, pancreatitis probably being the most famous example of sterile SIRS, burns, some metabolic diseases, thyroid storm and adrenal insufficiency, a number of drugs can cause SIRS, a blood product reaction, the classic malignancy, renal cell cancer, tumor lysis syndrome, and even a subarachnoid hemorrhage would bleed into the ventricles. They'll all present in a similar fashion to sepsis. So it's a very difficult diagnosis to make. And I'm borrowing the quote from Jean-Louis Vincent who wrote an excellent editorial on this where he said, Dear SIRS, you are overly sensitive. And that's precisely what the problem is. And this is a classic study that was published in JAMA in 1995. Prospective cohort study with a number of ICUs and general hospital wards where they surveyed over 3,700 patients and actually 68% of the patients in the study at some point met the criteria for SIRS. So not very specific at all. And of those, only 26% ultimately were diagnosed with sepsis. So clearly from a screening perspective, SIRS, you'll end up with alarm fatigue, et cetera. You can't go with SIRS. Then as an infectious disease doctor, there's the problem of infection. This is the trying to actually make the infection component of the sepsis criteria is not easy because cultures take 24, sometimes as long as 72 hours to return positive. Oftentimes somebody's been slipped a couple doses of antibiotics at home before they came in or antibiotics before the cultures are taken and that decreases the yield of the cultures. Although sepsis and bacteremia are often used in a synonymous, they're often thought to be synonymous, they're not. In fact, only about 30% of patients who have sepsis are actually bacteremic, actually have blood cultures positive. And the fourth bullet point really picks out the issue of culture negative sepsis. And in this very large study of over close to 7 million patients, 47% of patients with sepsis were culture negative. And of course we all know with the bundles, the real pressure to give antibiotics within one hour, imagine what this does in terms of the excessive use of antibiotics. We're giving lots and lots of antibiotics to people who actually don't have sepsis. So then finally in 2016 we got away from the Las Vegas criteria and moved on to CEP3 definition. And it's nice because it no longer describes an overwhelming systemic inflammation, but it describes a dysregulated host response to infection. So you could actually have excessive inflammation or you actually could have immunoparalysis both in the setting of infection. And for that diagnosis you need to have at least an increase in two points of your sequential organ failure assessment, your SOFA score, that's if you're in the ICU. And if you're out of the ICU, a QSOFA, a quick SOFA that goes up by two points or more, that is of two points or more, and the three components are listed, hypotension, altered mental status, and tachypnea. So that's the CEP3 definition. Now the problem is that the SOFA itself has not born, this new SOFA definition and QSOFA definition has not been born out to be that helpful either. So this is a study that was published in Chess by Anand and colleagues. QSOFA evaluated in more than a million patients at 85 hospitals. The sensitivity of the QSOFA was only 63%. So one out of the three people who had sepsis did not have a positive QSOFA score. So you were missing a lot of people. So you didn't have the sensitivity that you wanted. And in terms of positive predictive value, only one out of the six people who had a positive QSOFA score actually had sepsis. So you weren't getting the specificity either. So you really weren't making any headway in either direction. And so what you end up with is 83% of people with a positive QSOFA that were getting unnecessary antibiotics. So the new definition really hasn't helped us all that much either. So people have moved on, and this is not an exhaustive list, but they've moved on to can we come up with single biomarkers for sepsis that will help us beyond these clinical definitions. And in yellow are the markers you actually can get right now. Yellow with an asterisk is ones that require a special platform. So you can get an absolute neutrophil to absolute lymphocyte ratio, an absolute lymphocyte count, absolute eosinophil count, red cell distribution. You can get those all off your standard CBC. You don't need any special technology. CRP and ESR, almost all clinical labs have. Procalcitonin, you need a special platform. Immature granulocyte, you need a special analyzer. Same is true for monocyte distribution with. The endotoxin activity assay also requires a special platform. And then there's a whole host of biomarkers, and I apologize to anybody in the audience if their favorite biomarker is not in the list. But the problem is there's a lot of biomarkers. But at the end of the day, we really don't use many. We really don't have a lot. And one biomarker probably does not describe the complexity of sepsis. So a couple examples of the ones we use, right? So C-reactive protein, it's readily available. The problem is it has a long half-life. It stays up for days after trauma or surgery. It goes up in inflammation due to neoplastic conditions, as well as rheumatologic conditions. So it's not specific. Drug reactions, CRP also goes up. And in terms of a cutoff, a cutoff of about 90 milligrams per liter for CRP had a somewhat useful diagnostic ability in terms of its area under the receiver operated characteristic curve of .794 in distinguishing non-infected sepsis from sepsis. But certainly not ideal. Same thing with procalcitonin. It has a shorter half-life, so it doesn't stay up for as long after a major trauma or surgery. But it is up for a period of time. The nice thing with procalcitonin is it's not affected by drug reactions, neoplasms, or rheumatologic diseases. And at a cutoff of about one to two nanograms per ml, it has a better AUC to ROC curve of .85 for distinguishing between sepsis and SERS. So an improvement over CRP for sure, but not ideal. And there are things that can make your procalcitonin go up. Medullary thyroid cancers, more commonly small cell lung cancers, carcinoid tumors. A number of biologics, including ATG and rituximab. It does still go up after trauma and surgery. However, it doesn't stay up for as long. Hard to use in neonates because procalcitonin is elevated in the first couple days of life. And there is one rheumatologic disease, Stills' disease, where it is up. So there are confounders in the use of procalcitonin. So how could we move beyond single biomarkers to try to hone in on the diagnosis of sepsis? So a new approach is looking at gene expression. So it is an attempt to understand the sepsis response at the transcriptional level of mRNA. So going from the gene itself to RNA to protein. So this is at the RNA level. And it's accomplished by isolating RNA from leukocytes, hybridizing the RNA, putting it on a chip, being able to quantify RNA changes, usually with PCR, and try to align these changes into a signaling pathway analysis. So multiple genes being turned on in response to a stimulus. And one such attempt was the first iteration of what we're talking about tonight, which is the septicite lab. And the way this development proceeded was there was a derivation cohort of patients from five ICUs in Australia. The blood was collected in PAX gene RNA tubes so that you could isolate RNA. And they tried to differentiate 74 cases of sepsis from 31 cases of post-surgical, non-infectious inflammation. And what they arrived at was a gene expression signature that had four markers. Now, I won't get into the specifics of these markers, but all of these are involved in the immune response. And then what they attempted to do was to validate these markers by reverse quantitative PCR in five independent cohorts of ICU patients in the Netherlands, associated with the PCR. In the Netherlands, the so-called Mars Consortium. And what did they find? They found that the AUC to ROC for differentiating sepsis from non-infectious inflammation ranged between 0.85 and 0.95. So an improvement over what you saw with procalcitonin. And that was across all the validation cohorts. Another key finding in this paper by McHugh and colleagues was the ROC was not affected by severity of illness. So it didn't matter what Apache 4 strata you were in or what sulfasource strata you were in. When they compared the septicite lab to procalcitonin, consistently the AUC to ROC ratio, which is a combination of sensitivity and specificity, was 0.10 to 0.17 higher than procalcitonin across all the cohorts. And it also had a higher AUC to ROC curve for combinations of commonly used biomarkers in sepsis. And for me, one of the key points was the septicite score of less than 4 had a negative predictive value of 98.8%. And how is that important? Well, if you have a pretest probability that your patient has about a 30% chance of having sepsis, if you had a result below 4, that lowers your post-test probability to 1.2%. So a very good rule-out test. So the conclusions that were drawn from this early study were that the septicite lab differentiates sepsis from non-infectious systemic inflammation with a relatively high negative predictive value. It doesn't require that you identify an infecting organism, so it gets you away from the problems associated with the current state of microbiology. It's not affected by the disease severity. Its diagnostic utility goes beyond what's currently available, in terms of the markers and labs I've already showed you, but that it is going to require further studies. And ideally, and you're going to see a demonstration, it needs to be incorporated into a platform that is very user-friendly, that you can actually not need a lot of tech time, a lot of space, and you'll get an answer in a time that you can actually have actionable information. So Dr. Miller is going to get further into the development of this technology going forward. Thank you. Thank you, Dr. LaRosa. We didn't allow any time for question, but it's a fairly informal situation here. So if you want to go and ask him questions afterwards, he's made himself available. You can have another glass of wine, another plate of tappers, and grill him on some of his presentation. So next, we're going to have Dr. Russ Miller, who's going to speak. And he has been a collaborator with Immune Express for many years. And most importantly, he was part of a large group of collaborators in a study, the Venus study, which was taking the septicide technology and really kicking the tires, really trying to see if it could be validated with clinical samples. And he's currently at First Health of the Carolinas. When he was engaged with this collaborative study, he was at Intermargin Health. And he has also a glittering career and resume that you can see here. So I will let him take the podium and tell you about the Venus study. Thank you. The searching for the unknown is what I elected to entitle this talk, and the idea here that I think Dr. LaRosa has laid out pretty well is the concept that there's a lot unknown in discrimination of cases of actual substance, but also, frankly, in how we're going to measure ourselves and our care against whatever definition we come up with, which he spoke of also. So why do you care? Well, it's an interesting question, or perhaps, I hope, an interesting question, not just to say why does this matter to you, but why do you care? It is estimated, and I'm an intensive care doctor, it is estimated that somewhere around between a third and a half of us, by the time we get another 20, 30 years, hence, in age, will spend time in an ICU, and the reality thus is that all this matters to us. It's a lot better matter to us intimately, and we have an obligation to try to find solutions where there seem to be none, and so we know, as Dr. LaRosa has laid out, that it's very difficult to differentiate. An example he didn't give, but I think one that is also germane to us is AFib with RVR, right? Somebody comes in who just has some tachypnea from pulmonary edema, but also, of course, tachycardia associated with the AFib, what do you do with that person? And the truth is, of course, some of those are infected, and that's why they've gone into AFib with RVR. So we need a test. We need a test to begin to do this discrimination, and clearly, there are tests out there that begin to think about detecting organisms, but we need something novel, something that looks instead of the host side of the equation, and in deference to my infectious disease colleague, I've even included a nice epidemiologic term, which is we don't even have an alloyed standard in this syndrome, do we? We don't even know exactly how to make this determination. So what did we do? How did we do this? So two things. Dr. LaRosa has already described the work that had happened prior to my involvement with Immunexpress as well in developing, and then we did the process of evaluating and validating in a U.S. cohort, mostly U.S. cohort, although I'm going to show you some data both from U.S. and Netherlands patients, and the concept was to think of this as an early critical illness, if you will, so a patient perhaps in the emergency room or wherever they might be coming from, but they were being looked at in the very early stages of coming into the ICU. We studied ultimately, the numbers add up to 447 between these two cohorts, one from the U.S. and one from the Mars study. We called our study Venus because we were trying to be cute, and I think we succeeded. Another glass of wine, all of you will laugh. So the concept here was absolutely to get as much data and as many patients as we could and then go about using this lab that Dr. LaRosa described in a way that perhaps was a little different than he described. He initially described sort of a less than four, greater than four, for example, and what we did in the context of this prospective analysis was to generate a quantitative score that crossed four different bands, and the exact numbers are of no relevance to anyone in the room other than zero is the minimum, ten is the max, and the fact that when I show some data in coming slides, we will look at that across these same four bands. So that's how we did it. But how do we know if it was sepsis? This is the big question. And so there was a lot of discussion. I admit that there was never heated, but it was fun. I got picked on a number of times for things I said by various people, Dr. Davis included in the back, but in the process, we looked ultimately at three different definitions of sepsis. And so they are outlined here, and we're going to walk through them because I think it's important to understand what each of these is. So one, and all of these were independent experts, so this was not anybody assigned or it wasn't me, for example, in making this determination. We used what we called a forced model. And that forced model basically said you cannot have indeterminates. And so patients either had to be assigned into sepsis or SERS. Now, I think we can all appreciate you're taking the mud and throwing it into the water when you do that. And so if you will, this was thought to be a reasonable estimate, perhaps, of the lower bound of the performance of this test. Similarly, and I'm going to jump all the way to the end just to create the extreme, the unanimous model is where all three panelists agreed. Now that happened, although you can see it was not perfect, of course. And when they didn't, it's a thing, and when they didn't, you had a couple different scenarios. If everybody disagreed, which happened a few times, one would say sepsis, one would say SERS, one would say indeterminate, or if they all said it was indeterminate, which I don't personally remember the numbers, but I suspect was pretty uncommon, of course, we knew that we then assigned the diagnosis as indeterminate. Now that is now, again, leaving the mud in there, but as a separate category. So clear water, clear water, and a little mud. So the consensus model, then, was ultimately a model where we only required two panelists to agree. And the concept here was it's as good as we're going to get. And the other thing, if that sounds bad, the other thing is it's a, and this term gets overused, I'm going to use it anyway, bit of a real-world model, right? It's how would you and two of your closest colleagues come and look at this case? And if two people agreed, in many cases, that would be sufficient to proceed along one course of action. For example, treating sepsis and determining that it is, in fact, sepsis. So what did we find? And I'm going to go through each of these four points on the subsequent four slides, but first of all, we determined that septicite lab had a relatively high, in this field, shockingly high, one might argue, AUC, in the .82 to .89 range. And it did depend on the algorithm, and I'll show that, these different, three different models, if you will, or algorithms, the forced consensus and unanimous. And then that the relative likelihood of sepsis increased, so as this score, ranging from zero to 10, increased, the likelihood of sepsis also increased. It gets back to that notion of biologic plausibility, right, of any new test or any diagnostic. And then we did a regression analysis for looking at this concept of how does this perform relative to other clinical and perhaps laboratory-based markers. And finally, I'm not actually going to show these data other than a footnote, but it kind of comes back to one of Dr. LaRosa's points, which, in his case, was related to severity of illness not having an effect. Similarly, I will not show, but trust me, the demographic data did not play into the performance of the test. So let's talk about discrimination in the first case, in the AUC, here, and I have attempted to then, oh, it doesn't work on the screen. I have attempted to then look at each of the models, obviously, okay, and with each of these models, you can see a variance from low to high in the actual AUC. And the idea here, again, is that if you treat the top forced model as the lower bound and the bottom one as probably the highest possible bound, then the real world or consensus does represent, in fact, the middle. Again, some plausibility, I think, represented by this because there is that expected gradation in the correct direction of the results. Similarly, as the score goes up, and now you see, if you look just upon panel A, you see the forced model and you see four dots, those four dots represent the four bands that I referenced earlier. And you can see that each one of these takes a nice parabolic turn, right? And the idea is that certainly by the time you get up to band 4, certainly by that time, you are very well discriminating sepsis and SIRS. So that there is, again, a dose response effect, if you will, of this test. And then perhaps most importantly, and I think this has been alluded to also in some ways, how does this compare to what we might otherwise be stuck with at this point? So if you look at the gray, just the general clinical markers here, you can certainly see that the AUC predictably is terrible in that context, right? Predictably terrible. You can even see that how procalcitonin specifically, PCT here, added to that body of literature and to the evidence we have. And yet, very clearly here, SEPTA score pushed us even further to the right, further up the AUC, the desired AUC range, to a point that it becomes, I think, begins to approach obvious practice changing behavior. And the concept here was that even if you include, and you can read these 14 variables and the 16 trillion or 16,000, whatever the case was here, different permutations run to make this, to generate this figure, but basically the highest value or the highest predictive value came from using SEPTA score. So to put this in perhaps simpler terms, it looks like it does a pretty good job. I mean, AUCs of .85 in our world in critical care are almost unheard of. We just don't get there. And yet, this test hits that mark. Also and importantly, adding normal clinical and lab parameters or even procalcitonin to the mix does not appear to affect the performance. And again, I didn't show you the demographic data, but it holds in that context as well. And then finally, that it is an FDA approved test to, and the first of its kind, if you will, to look at the host response, to begin to think about sepsis differently. So great. What are we going to do about that? Well it turns out what I didn't tell you is relevant. What I didn't tell you is that all that study, I think it's the back right poster there, that study was conducted with a test with a turnaround time of four to six hours. Who wants to wait for those four to six hours to determine? First of all, you've missed half of the sepsis targets in that time, of course. You've got to make a decision faster. So if only septicide lab could provide results faster, gosh, there might actually be a market here. You're at a talk presuming there's a market. And indeed there is. Now interestingly, this slide almost hides the fact that perhaps I hope is obvious to you, and that is at the very bottom left, that this turnaround time on the new platform is thought to be somewhere around an hour. Now we're hitting the mark where all of a sudden this can change practice and change it quickly and change it efficiently for us. And more to come on that. Thank you, guys. Thank you.

sepsis

septicite lab

diagnosing sepsis

gene expression analysis

Venus study

turnaround time

Procedures

Infection

Systemic Inflammatory Response Syndrome SIRS

Presentation

Select

Professional

Associate

2020

Advanced

Sepsis Resources