Latest / CRITKNOW / The FAST trial: Rapid vs. Standard AST

Transcript

The FAST trial: Rapid vs. Standard AST · CRITKNOW

• Ramesh Venkataraman: Hi, hi everyone. ⁓ welcome back to our ⁓ Kritno Podcast. ⁓ you can follow us on YouTube, Spotify, and ⁓ Apple Podcasts. Also check out the blog written by my co-host Bharat on CCM evidence. I think condensed versions of statistical principles in a very easy to understand manner are there. So please check that out. And we usually put those links underneath our podcast too, so to make it easy for you. So Welcome back to our podcast. And Bharat, I'm thrilled to say that our subscribership is increasing. And ⁓ no, ⁓ I think it's encouraging that we should probably be doing more and more of this. ⁓ I thought today we will ⁓ discuss a paper that has ⁓ enough relevance and also has enough interesting statistical aspects to discuss. It's the fast trial that actually looked at rapid antibiotic susceptibility testing versus the standard testing and outcomes in in gram in patients with gram-negative bacteria. across several centers. So I thought that would be a good paper, ⁓ considering that it is relevant and has a lot of ⁓ nuances that we could probably discuss about. I think d the best way would as usual would be to go ⁓ go over the background and see where we are till now and then set in set the context of the paper and then go from there. So over to you Bharat. Bharath Kumar T V: Thank you, thank you so much, Ramesh. ⁓ Yeah, before we delve into the trial, as you said, it's been good to explore the background. And maybe a good view of looking at it would be to say, when a clinician sends a sample for blood culture, what actually happens in the microbiology lab normally or in conventionally? So in a modern microbiology lab, if you send a blood culture sample, ⁓ essentially the first step is ⁓ the blood culture bottle is introduced into a device that is constantly looking for microbial growth. So one of these systems like Bactec, which is essentially monitoring for carbon dioxide production. So that's the first step. And as soon as the machine detects carbon dioxide production, it sort of like gives an indicator that there is some growth in that bottle. The microbiologist then pulls out the bottle and then does two things. One, they immediately do a gram stain. And then obviously, they give you a result for the gram stain. They also plate it or subculture it. ⁓ And then once there is a sufficient growth of colonies, the next step they do is species identification. And they do this through technologies such as mild detolf in a modern laboratory, right? ⁓ Otherwise, they would do biochemical tests and so on in a conventional laboratory to test for what species it is. And once the species identification is done, the next step is the antimicrobial susceptibility testing. And this in a conventional lab would either involve disc diffusion or drop micro dilution or any of these techniques. Now this whole process, ⁓ even in a modern lab currently, takes anywhere between two to five days and it depends upon a number of factors. ⁓ But obviously as a clinician, we are on empirical antibiotics until then and we can either escalate or de-escalate only based on the sensitivity. So this process is actually pretty long drawn out and that's the context or that's the problem that the current fast trial is trying to address. That is one component of it, which is the antimicrobial susceptibility testing. So it's trying to test whether a rapid strategy of antimicrobial susceptibility testing would be superior to the conventional techniques that are already out there, right? ⁓ And in this particular case, they're using a technology called Vitek Reveal, ⁓ but There are, think, a few of these options available. And if I'm not wrong, there are also a few trials that have previously looked at this question. Yeah. Ramesh Venkataraman: Yeah. So ⁓ that's a very quick succinct way of looking at the whole thing. I want to just prompt our audience here that we do have a podcast that's going to follow this ⁓ with the microbiologist who's probably going to explain the traditional methods and the rapid methods in a little bit more detail because ⁓ I think we should go beyond the data as we say and just try to understand ⁓ what we do. In real life, and then interpret this paper in light of what we can do in our setup and then go from there. So, with that ⁓ introduction, I just want to quickly ⁓ review the literature so far, Bharat. So, what I thought is there have been about five trials looking at ⁓ a similar concept as rapid or fast anti ⁓ antibiotic susceptibility testing compared to the standard techniques, improve outcomes. And ⁓ this study has been done by Banner G et al. And there have been two previous studies by Banner G's team themselves. So they have been on it for almost now since 2015. So one trial in 2015, one trial in 2021. So there have been at least about five multi-centered trials looking at various populations and different sets of outcomes. The consistent message, rather than going into each one, consistent message is that. None of them have shown a mortality difference between a rapid antibiotic susceptibility versus standard. What they have, however, shown is that time to appropriate optimization of therapy. So target up targeted or optimization of therapy is actually faster when we do antibiotic susceptibility. Either if you started off with a wrong empiric antibiotic, your time to the first appropriate antibiotic is faster, or two. If you think you should be able to, you you think you can de-escalate and the time to de-escalation has been faster. So that has been a consistent signal across all the studies. So that's where we stand. And most of the other studies have been done in areas with low resistance. And and and so they thought that no, I I I I think that ⁓ this technology would benefit when the resistance rates are quite high. If the yield is too low, then maybe it's difficult to show outcome benefits in. Bharath Kumar T V: Mm. Resistance. Ramesh Venkataraman: I the event rate is so low. So I think this study mainly wanted to focus on high-income countries with high rates of antibiotic ⁓ resistance. So they have included Greece, one center from India, and and and other countries similar to our setup where they have looked at ⁓ patients with gram-negative bacteria and whether this technology improves outcome. So that's where we are at this point of time. Bharath Kumar T V: Yeah, yeah, I think as you rightly pointed out, this question sort of assumes more relevance and importance in contexts such as ours, right, where resistance rates are high and particularly resistance to carbapenems, which is sort of our biggest problem. So yeah, in this trial, know, yeah. Ramesh Venkataraman: One other point I thought we should add here is that the question always arises, is there a good agreement between these rapid tests and the final susceptibility testing, right? As a clinician, I would say I will get the result faster, but if I get it wrong, I'm going to s I'm going to not treat the patient well. The previous studies have a good track record and show almost like a ninety-four percent agreement between the rapid test and the standard test. So I think it's worthwhile knowing that they have been validated well before we go on to the study. Bharath Kumar T V: Mm. Yeah, absolutely. That's a very important point. So coming to FAST itself, they have included the population that the trial included is patients that have gram-negative bacteria and blood cultures, and they need to be hospitalized. They excluded those that were deceased by the time of the blood culture result if they had non-GNB organisms in the blood culture, if they'd been previously enrolled in the trial, or if there's a previous GNB in the blood culture in the past seven days. So I think reasonable exclusion criteria. The intervention, you were pointing out, is a rapid AST technology, which in this case was Vitek Reveal. And they do this in addition to the standard AST ⁓ in the intervention arm. And the species identification in the entire trial is through MALDITOF. The comparator was standard AST. And they left this to each of the labs, because each of the labs actually in the countries had their own ways of doing a standard AST. They randomized ⁓ 1S to 1 stratified by the site or the hospital as well as the ICU status. And they use block randomization in permitted blocks of four. Interestingly, the screening and randomization, and I think appropriately, happens from the microbiology lab. As soon as the culture is species is identified or the culture is positive, they then randomize to either of these ⁓ approaches. And the trial was conducted across seven medical centers in four countries. Greece, India, Israel, and Spain. all of them have a reasonable, of course, India has a high burden, but I even the other countries, including Greece, have a reasonably high burden of multi-drug resistant organisms. So that's sort of why they did the trial in those countries. ⁓ And then, you know, with the PICO, the key question is the outcome. And I think it warrants some discussion. So the outcome that they're chosen as something that ⁓ many people may not be familiar with. It's called the Desirability of Outcome Ranking Score or a DOOR approach. And in this DOOR approach, they have chosen three components. But I think it's worth sort of delving into the components themselves and what DOOR is all about, right? Ramesh Venkataraman: So ⁓ let me just quickly then go over to the outcomes, Bharat. I think this is where I think the study gets a little interesting in terms of how we go about doing it. So before I go into the result, just looking at like the characteristics of the patient's enrolled, the average age was about seventy three. Bharath Kumar T V: Did we want to talk about the door itself first, just the outcome itself? Ramesh Venkataraman: I ⁓ okay, all right. So we can talk about the door itself first and then we'll go into the results. I think that's a that's a good way of Bharath Kumar T V: Yeah, maybe we'll talk a little bit about DOOR and the analytical approach. Yeah. Ramesh Venkataraman: Yeah, yeah, I agreed. Agreed. I agree. I think we just quickly talk about what door is, how it's looked at, and then we'll go on to the results later. I think that will be a better flow. So a door, like you said, is desirability of outcome ranking. So in when you look at outcomes in statistics, you can have categorical or quantitative. Categorical can be a straight dead or alive, yes or no, and that's just a binary outcome. So it it is either this or that. But what if you have more than two categories? For example blood group no ⁓ more than two categories so then you could either have them in a ranked manner or an unranked manner for example blood group can't be ranked so those kind of no ⁓ variables that are more than two that cannot be ranked we would call them as a nominal outcome if there is some kind of a ranking then we generally call them ordinal outcomes ordinal outcomes do not have to have a specific ⁓ delta Fix a delta between them. So not necessarily like plus five or minus five. It's not like five, ten, fifteen. Doesn't have to be going like that. It doesn't have to have any particular sequence. So in this study, they have used an ordinal outcome where they have picked three categories. Anything more than two categories will qualify. So the first category obviously has got the top rank, is alive with no deleterious events, which means you did fine. Second, Bharath Kumar T V: Yeah, yeah, yeah. Ramesh Venkataraman: Is going to be alive, you had at least one or more deleterious events. And third is obviously death. So here it's very clear that there are three categories, and there is a ranking system, one, two, and three. And and our desirable outcome would be number one first, if that didn't happen, number two, and then the last outcome that we would desire would be number three. So that way, no, if you now think about desirability of outcome ranking, it kind of makes sense using ordinal outcome. So That's what they have done. So, what are the disarrable events that they have had? One is unsuccessful discharge. If you discharge somebody and they get readmitted within 30 days, that's considered unde unsuccessful discharge. That's a that's a deleterious event. Lack of clinical response, ⁓ either that they ⁓ no did not improve, develop new complications, whatever, new kidney injury, or While they are getting treated, they get colonized with a multidrug resistant pathogen. All of those were considered as relaterious events, and that's how the ordinal scale had improved. I am not a big expert in DOES, but I I think I'm going to also hand it over to you. From what I understand, DOR kind of looks at both the pluses and minuses of an outcome, right? I mean, here you have a desirable outcome and an undesirable outcome. For example, you can be alive. But you can have a deleterious event. So it looks at both the risk and benefit from a patient point of view, and you can get a holistic patient-centric outcome rather than just dead or alive. I think that becomes ⁓ very important. It can capture subtle nuances, right? Now ⁓ the the deleterious events can bring up a little bit more information than a traditional outcome can bring about. So it does capture some nuances and from a power analysis point of view than a traditional. Bharath Kumar T V: Yeah. Ramesh Venkataraman: Way of looking at it, it enhances the power with less numbers. I can get the same, it increases the efficiency of the way I do the study, and therefore my power power goes up. Of course, there are some downsides to the door outcomes. For example, in this case, deleterious events can be subjective, right? I mean, I can ⁓ it can my readmissions can be subjective, my complications can be subjective. So there is a little bit of subjectivity involved. Interpretation is difficult as as ⁓ I told you earlier, and as we will discuss, it's not as straightforward as interpreting a regular randomized control trial and therefore can be difficult for people. And there is a potential for masking very severe events by just clubbing all of them into more than one severe event, and we may not see if I have three or four catastrophic events, but not death, they'll all be clubbed together into one category versus somebody having a minor. No, some event also would be c ⁓ classified into that category. So there are some downsides of door outcomes, but this is generally the concept of door outcome, door outcome. And it's also interesting the way they analyze the data, Bharat. So ⁓ I think it gets a little more sophisticated. I think it'll be worthwhile I I think for us to go through how the data is analyzed compared to the traditional data. And ⁓ no, obviously you're welcome to add anything more on the door outcomes than what I have said. Bharath Kumar T V: science tool. No, no, that's I think an excellent summary of the ordinary outcomes, but also specifically in DOOR. The analytical strategy is also interesting. So in some ways, this is an ordinary outcome, but also in some ways, it's an ordinary composite outcome. And there is obviously an inbuilt hierarchy into it. So all of that would remind some of our listeners of the wind ratio discussion that we had in our very first podcast episode. So there are some similarities to the wind ratio approach here. The difference though is in the wind ratio, the ties were being ignored. Here even the ties are actually accounted for and I'll explain very briefly how. So essentially again here, the starting point is every patient in the intervention arm is compared to every patient in the control arm. And then they are looked, each patient is assessed for each of those events, alive without deleterious events, alive with deleterious events or death. And alive without deleterious events has the lowest rank, one. And then alive with deleterious events would have a rank of two and death would have a rank of three. Now for each comparison, ⁓ the ranking would then be compared. What ranks do each of the patients in those comparisons achieve? Obviously the participant with the lower rank is the winner, right, for that comparison. So they look at the number of wins in each comparison and the number of ties. And then they use a formula to calculate the probability that a given patient in the intervention arm will do better than a given patient in the control arm. And in this particular trial, they had defined or they used a probability of 55%. If the probability is greater than 55 % as something that would be meaningful and they had put this into their sample size calculations. Now it's debatable whether 55 % is a great, you know, improvement in terms of the probability. Obviously, if the probability is 50%, it's considered as no difference. And they said if the confidence interval of their estimate, ⁓ if that, if the lower limit fell below 50, again, it would be non-significant. But the 55 % itself, whether it's the right number, we'll again talk about it a little bit more. But this is the approach that they ⁓ took to estimating the probability. Ramesh Venkataraman: To interrupt you, they also inbuilt a tiebreaker, right? If two people end up being dead, then they both get the same rank, and then they looked at the time to death, who died faster would get a higher rank compared to a lower rank. If b both people belong to the same category, say alive with deleterious events or alive without deleterious events, then they looked at hospital length of stay or duration of stay as the Bharath Kumar T V: Correct. Correct. Correct. Ramesh Venkataraman: ⁓ tiebreaker. Whoever had the lesser hospital length of stay would get a better rank. So they also had some inbuilt tiebreaker. Curious Bharat, I mean, why why not a wind ratio? Why a probability? I mean when would you decide how how you would use these? Because that that study and this look like they are very similar, like you said, but they have used a different way of looking at the data. Bharath Kumar T V: Correct. Yeah. Yeah, I think ⁓ the main advantage here is that, you know, so they could have used the wind ratio approach here as well. There's no doubt about that. I think the advantage that is perceived is that you are accounting for the ties as well. You're not losing information by ignoring the ties, which we actually pointed out as a limitation of the wind ratio approach. So I think that's really probably what drives it. And in the ID literature, is now becoming a very common sort of an outcome approach. And many of the previous studies that used DOOR have used a similar approach of estimating the probabilities. It's also important to note that when they estimated the probability, they're doing a one-sided test. They're actually testing only for superiority, which again is debatable, right? Because any of these approaches can potentially cause harm. So they're only powered for a one-sided test. So that, think, is a bit of a limitation. But that's sort of what they have done. And no, I don't think there is a at least a statistical And so to say if one is better than the other, but the fact that there's a course for ties, I think helps, you know, with interpretation of it. Ramesh Venkataraman: Yeah. I think that makes sense. So just ⁓ I think these two things make the study very interesting. Once the door outcome, like you said, most of the ID studies now look at DOR outcomes and the way they have analyzed it becomes important. Now, quickly going on to the results. I think the baseline characteristics-wise, the median age was around 73, so slightly elder population. What I would like to point out is only about 11% variation population. Predominantly it is white population. Not that they would behave differently, but something to keep in mind. 10% of the patients roughly came from India from one center. I think it was Manipal. And about 17% of the patients were in the intensive care. Mind you, these are all patients who are hospitalized, not necessarily in the intensive care. So roughly one-fifth of the patients were in the intensive care. And about forty to forty-five percent of the patient had resistant bugs, either to cephalosporins or to carbapenems, and about a quarter of them needed wasoppressors, roughly twenty-four to 30% of the patients needed wasoppresses. So a reasonable population, both for a hospitalist point of view, from an intensive point of view, I think there is enough representation of the ICU patients where we can say, you know, that this is probably worthwhile for us to analyze. 55% had hospital-acquired infections. So again, that's something that ⁓ we can ⁓ you know relate to. And and ⁓ they also give a list of what infections cause the gram negative bacteria, which looks pretty similar to. What we would see. So those were the things. When we look at the primary outcome in terms of the ranking that we talked about, if you look at all the three, alive without deleterious events, alive with one deleterious event or death, there was no difference between the between the groups across any of those. The overall probability that ⁓ the rapid approach was useful as only 48%. So clearly fell short of that 55%. And the confidence interval was anywhere from 45 to 52 percent. And just to remind the audience, this study would be considered positive only if the lower limit of the confidence interval of the probability is at least 55 or above. Here we got 45 to 52. So clearly the primary endpoint for superiority in terms of the rapid antibiotic susceptibility was negative. So there was no difference. But we can't just throw this off. Right there because we need to look at a lot of different things. So, one, there is no difference in 30-day mortality between the two groups, roughly the same. There's no ⁓ difference in the eight days in terms of how they behaved. So, again, no acute ⁓ changes in outcome or chronic changes in outcome. They have given a list of secondary outcomes. For the sake of time, I'm going to jump to the two or three that actually matter to us. What happened to the subgroup of patients who had cephalosporin resistance or Carbipenum resistance. For cepillosporin ⁓ resistance, day to days to discharge was 8 versus 11 days, meaning that we are able to discharge the patient a little bit faster. Again, confidence interval was across zero. But when you look at escalation or de-escalation, if you had started an inappropriate antibiotic, about 60% of the patients in both groups had received appropriate antibiotics to start with. So about 35% did not receive appropriate antibiotics. So Either the time to escalate to an appropriate antibiotic, or if you're already on an overkill antibiotic and you want to de-escalate, the time to do this antibiotic sewer shift was much shorter in the cephalosporin group, 12 versus 30 35 hours, roughly about a 24-hour advantage, with confidence interval clearly favoring the rapid testing. And if you look at the carbapenem, hours to effective treatment within three days. Was 9.5 versus 28. Again, a whole day. You are faster by a whole day with good amount of ⁓ confidence interval favoring the AST. Again, hours to escalation and de-escalation was much shorter, 13 versus 47 hours. So roughly about a 36 hour, 34 hour advantage in terms of using the antibiotic susceptibility testing. And they also looked at ⁓ time to death. Or time to achieve this antibiotic therapy using Kaplan Meyer curves, and that's something that I think we should look for and say the group that got the faster antibiotic susceptibility testing got to faster escalation and faster de-escalation, and that becomes important. In the carbopenem group, the proportion of patients who were still hospitalized at 30 days was very low in the faster antibiotic susceptibility group compared to the standard group. Only 15% from the rapid group was still in the hospital compared to 28% with confidence intervals of minus 25 to minus 0.2. So clearly saying that there is maybe a potential advantage in that subgroup of patients. We can probably get the antibiotic right, get them out of the ICU and hospital faster. But mind you, they are all pre-hoc decided on subgroups. The primary outcome was negative, but I think. I I dwelled into the data of the subgroups purely because that is the subgroup that we would be interested in in our population, considering that the carbopenum resistance rates are quite high in our country and and our benefit if at all is going to be rep ⁓ you know replicate of what we see in that particular subgroup Bharat. Bharath Kumar T V: Yeah, no, I agree with you. And maybe a couple of points that I wanted to just add ⁓ on the results itself. one, the carbapenem resistance, ⁓ the proportion of patients is about 15 % if I'm not wrong, right? In the entire trial that were carbapenem resistant. Yeah. Ramesh Venkataraman: Yeah, yeah, yeah, yeah. Bharath Kumar T V: roughly about 15%. So I mean, again, for the listeners and viewers, there are some differences in that ⁓ subgroup. But you have to note that it's a very small subgroup. So the interpretation, again, has to be very cautious. That's something to note. ⁓ The other point that I think in the results that came out is that there were more stewardship recommendations in the rapid AST group ⁓ and more acceptance of those recommendations in the rapid AST group as well. ⁓ And the maximum patients, I think, enrolled from Greece and Israel. As I said, only about 10 % from India, and again, a smaller percentage from Spain. But I guess in summary, no difference in any of the key outcomes or in the components of those deleterious events. But there are some differences in terms of time to escalation, de-escalation, recommendations, and so on. That's the top line summary of the results. Correct? Yeah. Ramesh Venkataraman: Yes, agreed upon. Com agreed completely. Yeah. Bharath Kumar T V: Yeah, yeah, no. So I mean, maybe we should talk a little bit now about ⁓ strengths and sort of the concerns ⁓ from our perspective. So I thought ⁓ the first thing to say, ⁓ this is an important trial. And often when we do trials of diagnostic sort of techniques or technologies, people stop at just saying, what is the sensitivity, specificity, what is the diagnostic performance of the test? They rarely go on to check the clinical usefulness in a RCT like this. So I think it's very cool that they are doing this because that's what ultimately matters to us. So that I thought was an important strength. The methods overall, I think, were robust. The only concern that I had with the method was obviously the doors, the probabilities that they had chosen for their sample size estimation, which I had brought out earlier. I have a number of other concerns. The first thing is that Ramesh Venkataraman: Matters. Yeah. Bharath Kumar T V: There is no clearly defined MCID or minimally clinical importance for the door. How much is actually meaningful in terms of a door outcome improvement? think there's some work to be... Probability is how much is good, right? So it's 55 % good, 65 or 70. And there are some trials in the past that have gone to 65 or 70 as well, right? So that is something that I thought was a problem. Ramesh Venkataraman: How much probability of success you mean, right? ⁓ Or should we go to sixty five, seventy? Yeah. Bharath Kumar T V: And then you already pointed this out that some of the deleterious events are subjective. And the other problem here was that the people assessing for those deleterious events were not blinded. So if they were at least blinded to that, that would have been a way of mitigating the fact that they were subjective, actually some of them. ⁓ And the other point. ⁓ Ramesh Venkataraman: So so the just to just to ⁓ clarify that and expand on that, that stewardship antibiotic stewardship team, the entire lab team and people who assess the outcome, they were all unmasked to the randomization. So there is a potential for subjectivity in how the deleterious events have been defined. And since they have been used, then the whole door comp door components become a little bit subjective in terms of ⁓ how it they were ⁓ agreed upon. Bharath Kumar T V: Yeah. Yeah. Yeah, yeah, like some of the components are objective, but some of them can, you know, be subject to clinician. Yeah. So I think that was a concern that I had. And ⁓ although there is this finding that we, you know, you highlighted that the time to escalation or de-escalation is different or rapid for cephalosporins or carbapenems. Ramesh Venkataraman: Be subjective. Bharath Kumar T V: it's not making a difference to patient endpoints, right? Like the fact that overall it didn't make a difference. Yes, there's probably a signal in the carvapin group, but that needs to be evaluated formally, explored more. And if, mean, with escalation, obviously I'm looking for improvements. If the time to escalation was faster, right? If I was on an inappropriate antibiotic and I escalated faster, I expect improvements in cynical outcomes. Now with time to de-escalation, I would also look for Ramesh Venkataraman: Explode, explored more. Bharath Kumar T V: the overall ecological impact of what happens to the unit and the hospital's anti-biogram and so on. But obviously that's not something that the trial has looked at. So that would be of interest to me. So those are some of the things that came to my mind. What about you Ramesh? What did you think the strengths and... Ramesh Venkataraman: No, I I agree with all your points. For example, to to again elaborate on the ecology and ⁓ they did capture the occurrence of MDR pathogens while the patients were on this trial because that was counted as a deleterious event. But a as we discussed earlier, it is one among the many deleterious events and it can be sweeped across. We will not exactly know what happened, and I I I don't think that I think that's a very valid point, right? So we need to look at the ecological pressure. Bharath Kumar T V: Mm. Ramesh Venkataraman: In terms of what we do with the earlier escalation or earlier de-escalation, more importantly. I also had a few other thoughts, ⁓ final thoughts about this paper. I agree with the what the limitations that you brought about. ⁓ and if you look at the ⁓ like roughly about 450 and 450 patients, about 900 patients were generally originally randomized. Post-randomization in the rapid group, there were more exclusions after randomization than the standard group. It was 37 patients. Bharath Kumar T V: Mm. Hmm. Ramesh Venkataraman: out of 450 in one side and 12 out of 450 in the other. That kind of concerns me because there is an unequal ex exclusion for various reasons, some valid and some could be just no other reasons. That could confound the results and cause an attrition bias, right? You're you're censoring those events. So we really don't know whether that would have tilted the stud study results one way or the other. So that's something that we need to think for. And my take on all this is if if you think about a this test, If you have very, very little resistance, the utility of the test is useless. If you only have about 5% of the patients with carbapenem resistance, then no big deal in trying to know that you may miss out on one or two patients in a hundred, but you're not going to harm most of the patients with your empiric therapy, right? On the other hand, if you take a country like India, all of us are expecting every bug to be carbopenum resistant. Our initial antibiotic is always an overkill, right? We always cover for carbapenum resistant empirically because that's Bharath Kumar T V: Hmm. Ramesh Venkataraman: Kind of the ecology that we are there. And again, the utility of this test is going to be negated there also because everybody would get appropriate antibiotic early on because the high incidence of carbipenin. If it is very low, everybody would have gotten appropriate anyway. So both sides are gone. So I think looking at outcome differences with this test has to be very, very carefully done in the middle population, like anything else, that that U curve. So that is something. And number two. You have to understand that patients first need to flag the bacteria positive. So that would have taken 24 to 36 hours after the blood culture. And the randomization has happened within 16 hours of flagging the blood culture, which means patient would have been on appropriate or inappropriate antibiotic already for two to three days. So any time you you you you upgrade the antibiotic to an appropriate one after that to have an impact on outcome is going to be unlikely. So Bharath Kumar T V: Correct. Correct. Ramesh Venkataraman: To me, the argument of trying to get the appropriate antibiotic early to change outcomes going to be unlikely because you've already lost the boat, because it's already gone two, three days, right? So where I see the maximum benefit I would get in my ICU is can I de-escalate faster? Somebody's on an aphrotoxic drug like ⁓ polymexin or a cholistin for a carbipinum resistance, and I find patient yes or nay, and then I can de-escalate it faster, and that would help me Bharath Kumar T V: Yeah. Ramesh Venkataraman: ⁓ in my ICU because the incidence of carbapenem is so high, I would have anyway covered for them. Can I de-escalate faster? And this study, at least in the subgroup, tells me that I can de-escalate maybe 36 hours earlier. Maybe a dose and a half of polymaxin can be saved. Now, the question that you need to think about is: is it worth putting through all blood cultures, investing in a system? Bharath Kumar T V: Yeah. Ramesh Venkataraman: That would ⁓ pick these up a lot more earlier for the sake of 36 hours of de-escalation ⁓ is a call that each institution and ICU has to take abroad. But I think we need to understand where we would get maximum benefit ⁓ out of a rapid antibiotic testing. I feel it is more in de-escalation than anything else. The other reason I feel the outcomes in terms of survival was not good is almost 20% of the bugs that grew were off-panel bugs. Bharath Kumar T V: Mmm, yeah. Ramesh Venkataraman: You have to understand this this rapid antibiotic testing has a panel of bugs, right? So if you get a bug that is outside this panel, let's say you get a weird bug. In India, they got a few salmonella, which was not there in the antibiotic panel. Then after all that, you don't get the information that you need. You still have to wait for the standard therapy. So that could have hurt the investigators in terms of getting a higher probability, I felt. And the last thing is the turnaround time for standard. Bharath Kumar T V: Correct. Ramesh Venkataraman: Culture techniques itself is variable across all these five countries that have looked at, right? Some my hospital standard turnaround time may be different than another hospital. And the time lag or the time advantage I get from rapid antibiotic testing will be variable across all these centers. And that heterogeneity could have diluted some of the results that ⁓ we have we have had also. And the authors authors also point out that there was a shortage of. Bharath Kumar T V: Yeah. Ramesh Venkataraman: astrionam in one or two of those centers that even if you identified a carbapenum resistant and you wanted to add astrionam you weren't you would you didn't have. So the outcome benefits are contingent upon you picking up the antibiotic resistance, having good antibiotic stewardship, accepting their recommendations, and having the availability of the appropriate drugs all have to line for you to get into an outcome. So I felt it is unfair to ding the investigators to say no outcome benefit and move on because there are so Bharath Kumar T V: Mm-hmm. having the right drug, yeah. outcome. Ramesh Venkataraman: So much of those complexities that are involved. I am enthused by the fact that in the small portion of patients with carbopenum resistance, there seems to be a signal. Maybe the next study has to be looking only at those patient populations. I don't know how to design such a trial, maybe even more tricky and will take longer. So maybe that is a subgroup ⁓ I would take. If such a study doesn't exist. My answer is can I de-escalate it faster? Is it worthwhile for me to invest in this technology for that 36 hours of de-escalation? Those are the questions that I would answer. Rather than saying fast ⁓ the antibiotic accessibility testing is good or bad, I'm going to say we need to be aware of where we are testing it, what setup we are testing it, who we are testing it, how is it being used, and am I have access to the right antibiotics? If you do all of those, Bharath Kumar T V: Yeah. Yeah. Ramesh Venkataraman: Maybe it'll have more value and if the link is not completed anywhere, then having a test alone is probably not going to be of added value. That would be my take, but Bharath Kumar T V: Outcomes will change. Yeah, no, I think that's a great summary. ⁓ My own takeaway is that ⁓ I think that as a general rule, this is probably not yet ready for prime time, this technology. ⁓ The carbapenem-resistant group or the cephalosporin-resistant group is where there's probably value for money. But ⁓ my view is that it needs to be rigorously tested ⁓ in that population, right? ⁓ So. Ramesh Venkataraman: Sha. Bharath Kumar T V: Those actually were my final takeaways that it's an important trial, but I don't think that technology can be rolled out just for everybody that needs a faster AST. Any other takeaways, Ramesh? Ramesh Venkataraman: No, no, no, but I I think this trial was fascinating to read because of the door outcomes and the way they have ⁓ probability, the way new unique way of looking at the outcome. So that's a new learning point for me and I'm sure sure for the audience. And I'm curious to see what our microbiologist is going to say in the next podcast in terms of how ⁓ a microbiologist interprets these results and their and her perspectives is something that I look forward to hearing and I encourage our audience to also know check on that podcast right after this. So thank you. Bharath Kumar T V: Yeah. Absolutely. Absolutely, absolutely. Thank you all for ⁓ joining and watching the session. And ⁓ as a reminder, please like, comment, subscribe. We appreciate all of the feedback that we get. Thank you. Thank you all so much. Ramesh Venkataraman: Absolutely. Thank you.