Latest / CRITKNOW / SOHO trial: HFNC vs. Standard Oxgen Therapy
Transcript
- Bharath Kumar T V: Hello everyone and welcome back to Crit-No. Ramesh and I are very excited today for a couple of reasons. One, course, for the trial, but also because we have our Crit-No T-shirts now and we are pretty excited about it. A couple of quick announcements before we get started off. Our podcast, as you know, is available on Apple, Spotify, and also on YouTube. If you've been liking what we've been doing, please do subscribe to our channel and also add some comments and let us know what you think. We are very grateful for all the feedback that we've been receiving. so far. The second announcement is that we are supplementing what we do here with a blog called ccmevidence.blog where we take a deep dive into some of the nuances that we discuss on the podcast. So do check out ccmevidence.blog as well. Today's trial is the SOHO trial, the trial that compared high flow nasal cannula versus standard oxygen therapy in patients with acute hypoxemic respiratory failure published in the New England Journal of Medicine just over a month back. ⁓ I'm going to now pass on to Ramesh to let us know a little bit about the background to this trial. Over to you Ramesh. Ramesh Venkataraman: Thanks Bharath. So the SOHO trial looked at high flow nasal canella versus standard oxygen therapy. Before we go into the context of the trial, I thought I'll just briefly take a minute or so to discuss the physiological rationale of high flow nasal canella. So as the name suggests, it is a high flow system where the minimum flow usually is at least 40 to 50 liters most often. The standard oxygen delivery devices usually go up to 15, 1, 5 litres which is nowhere close to the 5, 0, 50 litres that are higher that HFNC can actually good. So, one because of the high flow the oxygen delivery is relatively constant and quite high compared to what is achievable with standard oxygen therapy. Second, the high flow causes some amount of a peep effect with an average peep of around 3 to 5 which could help. the alveoli to be partially open and this can decrease your inspiratory effort and unload your inspiratory muscles on top of it. The biggest advantages that we have all seen visibly with HFNC during especially during the COVID era is its extreme ⁓ comfort for the patients. Patients seem to tolerate it well because of the high amounts of humidity they get compared to somebody getting standard oxygen therapy where Humidification may not be optimal and they just get dry air blown through their upper airway. So all these reasons have slowly pushed high flow nasal canola into the practice of critical care medicine and it's gone. It's ⁓ become one of the first to go devices for acute respiratory failure, especially hypoxic respiratory failure. So that's how I think HFNC ⁓ started infiltrating into our practice. Over to you Bharath, could you just basically maybe summarize the key evidence so far comparing HFNC with maybe standard oxygen delivery device. Bharath Kumar T V: Yeah. I think one of the first major trials to evaluate HFNC, which I'm sure listeners and viewers would be familiar with, is the FLORALI trial. This was published over 10 years back now in 2015, a French trial from the same group that conducted SOHO trial as well. So this trial included approximately 300 patients and the trial compared HFNC to standard oxygen as well as NIV. arm trial in acute hypoxemic respiratory failure. Their primary outcome was the need for intubation by age 28 and mortality was a secondary endpoint. They demonstrated a lower risk of intubation in HFNC as compared to the other two arms, but this did not achieve statistical significance. And they had a post-hoc subgroup where they looked at patients with a PF ratio less than 200, where the difference in the rates of intubation was also statistically significant in failure of the HFNC arm. But there was of course a post-hoc post hoc comparison. There was no, in the 90 day mortality, there was, ⁓ you know, it appeared to be lesser for the HFNC arm as compared to the standard oxygen arm and the NIV arm. But obviously the trial was not really powered for it. It was a small trial and the primary outcome was the intubation rates. Now, ⁓ after this trial, we've had the high trial in 2018, which was a trial that was conducted again in France, which focused only on immunocompromises. patients and I think they enrolled close to 700 patients which compared HFNC to standard oxygen therapy with the primary outcome being 28-day mortality and they did not show a difference and in their trial did not show a difference in intubation rates as well. Then there was a network meta-analysis published in JAMA in 2020 which included over 3800 patients and the meta-analysis demonstrated that HFNC had a lower risk of intubation but no difference in mortality. Then we have a couple of relatively more recent trials, the SOHO COVID trial, which is actually the same group, but in COVID patients in 2022, which compared HFNC to standard oxygen therapy against 700 odd patients. No difference in the primary outcome of mortality, but incubation rates were lower with HFNC, 45 % versus 53 % with standard oxygen. And then we have the RENOVAID trial, which was published last year in JAMA, which was actually a sort of a larger trial of 1,700 patients, but comparing asking a slightly different question asking whether HFNC was non inferior to NIV and they had five groups of patients in the trial so they had the immunocompromised hypoxemic respiratory failure the non-immunocompromised with hypoxic failure they had a acute cardiogenic pulmonary edema group they had a COPD group and then they had a COVID group and they ⁓ depending on the know the analytical model they use they claim non-inferiority most of the groups except the immunocompromised group but they some questions around the non-inferiority margin chosen and the analytical approach. I mean the summary of the evidence is that it appears that from previous data HFNC seems to reduce the risk of intubation but no clear benefit on mortality. Ramesh Venkataraman: No, I think that's a great summary of all the key trials that you have hit hard there. ⁓ I think the rationale for this study comes from that. think people tend to feel that Florali showed a mortality benefit and then the subsequent studies didn't show but various, they had the different population, different kinds of interventions were done. So they thought we needed a definite study looking, comparing HFNC plainly with standard oxygen considering that is increasing penetration of HFNC into our practice right now. So with that, I think the SOHO trial picked patients about 1,100 patients about 550 odd in each arm randomized to either high flow nasal canola or standard oxygen therapy. The inclusion criteria are pure acute hypoxic respiratory failure with the respiratory rate more than 25, a PF ratio of less than 200. and at an oxygen of at least 10 liters per minute or more through an NRBM. So, if you meet those criteria then you get randomized to either HFNC or standard oxygen arm. If you are in the HFNC arm, they started you off with 50 liters flow and then they titrated the FIO2 to get a saturation of 92 to 96 which seems reasonable. In the standard oxygen arm, they started off with a 10 liter oxygen flow. and then titrated the FiO2 to again get a saturation of 92 to 96. I think patients in the HFNC arm got HFNC for at least two days and after that the HFNC was weaned off if these criteria were not met, meaning if the respirator rate had dropped down to less than 25 or the PF ratio had improved. The exclusion criteria I thought were pretty standard. They said they excluded acute hypercapnic respiratory failure. There's always the worry that HFNC can make hypercapnia worse. They eliminated patients with acute COPD, patients, chronic lung disease, acute cardiogenic pulmonary edema, where the first modality would be noninvasive ventilation, hemodynamically extremely unstable patient with an R-epinephrine dose of more than 0.3 mikes per kilo per minute. So these are some of the top exclusion criteria, which I thought were very, very valid. The two things struck me as unusual in this trial. One is that they discourage the rescue use of non-invasive ventilation. That is not how you and me would be practicing care at the bedside. And the authors don't give a good rationale. Maybe they just wanted to keep their groups very clean and maybe don't know crossover with NIV. That could be the only rationale I can think of. Second is there is no data safety monitoring board in this particular trial. Most major intervention trials nowadays mandate a data safety monitoring. board but the authors have clearly mentioned that according to French regulations they met criteria for exemption of DSMB in this trial. So we will let that go at this point of time. The primary outcome was 28-day mortality in the study. That is the primary outcome. They looked at several secondary outcomes. The most important secondary outcome they looked at is the need for intubation in both groups and the next one would be the median time to intubation. ⁓ and then ventilator free days and then so on. They looked at improvement in PF ratios and so on subsequently. But I think the three relevant outcomes for me, three or four relevant outcomes are mortality, which is the primary outcome, need to intubation, time to intubation and ventilator free days while they are in the ICU in the first 28 days. So that's kind of how the trial was done in this particular ⁓ SOHO study. ⁓ Quickly going over Bharat, could you just tell us what kind of patients they enrolled and what actually happened to those patients? Bharath Kumar T V: Yeah. ⁓ I think as I said, very standard inclusion, exclusion, and I think pretty good, ⁓ I guess, guidance to sites in terms of how they need to use the HFNC in standard oxygen therapy. think that was sort of important. ⁓ They also had, I think, criteria for intubation, right, for when patients could be intubated. Of course, was still the initial discretion built into that. So they enrolled ⁓ over 1,000 patients. The mean age of the participants was 62, and predominantly male and you know close to 65 to 70 percent of male participants. The mean SAP score was 35 which is sort of a moderate SAP score and the median so far at the time of randomization was 2. About 20 percent of the patients or so were immunocompromised in the trial and the main reason for entry into the trial was pneumonia which is primarily viral pneumonia and then bacterial pneumonia. And coming to the primary outcome of mortality it was 14.6 % in both arms, so no difference at 28 days. Their most important secondary outcome, which is the need for intubation again by day 28, was 42.4 % in the HFNC arm versus 48.4 % in the standard oxygen arm, so a 6 % reduction in the need for intubation. There was no difference in the time to intubation. There was no difference by any subgroup and their sensitivity analysis. sort of confirm their primary ⁓ approach. In terms of adverse events, there were slightly more pneumothorax in the HFNC group, but overall, you the numbers are small, so I'm not sure we can read too much into it. So in summary, know, no difference in 28-day mortality, a 6 % reduction in the need for incubation, no concerns in terms of the time to incubation. It appears to be similar. So Ramesh, what are your thoughts based on those results? Ramesh Venkataraman: Mm. Yeah, a couple of points Bharath. What I thought is the 50 % of the patients had COVID. So I thought that was skewed towards a COVID population. We don't see as many COVIDs anymore in the ICUs we practice. I do agree that viral pneumonia probably happen more often than bacterial pneumonia. But I'm quite skeptical I get 50 % COVID pneumonia at this time point of view. But it could be because they had included some patients from the previous Bharath Kumar T V: Mm-hmm. Ramesh Venkataraman: when the COVID enrollment were happening, I think those patients were included in this trial. The authors make a point there. Second thing I found was the PF ratio less than 200 means they are severely hypoxic and about one fifth of them had a PF ratio of less than 100. So quite sick. But if you look at the baseline characteristic, only 2 % had vasopressor use. So they are really not multi-organ failure kind of patients. They are single organ acute hypoxic respiratory failure kind of patients. I thought... These two stood out for me when I looked at my patient population and said, okay, does this trial apply to my patients? And then now coming to the outcomes. So the mortality I feel may not be the best outcome to look at a study like this, right? mean, HFLC is an oxygen delivery modality. And between the time we give oxygen and the endpoint of outcome, which is a very distal endpoint, there are several other care processes that happen. that can clearly confirm the mortality. For example, the need to intubation, how you ventilate them, ventilator-induced lung injury, appropriate antibiotics, steroids versus no steroids. I mean, I can go on proning versus no proning. I mean, you can keep on looking at all the other variables that can increase the mortality. So that's something ⁓ that I would like to mention. Second is, if you look at the Florali trial, I think the control group mortality in the ⁓ control group was much higher compared to what we saw here. Traditionally, as time goes by, mean, between Florali and SOHO, it's about a decade. So as time goes by, the mortality of most diseases drops. And therefore, I think new trials kind of shoot themselves in the leg by trying to use old data to power their trial. And therefore, they find that the control group now has less mortality. And then to show a mortality difference becomes quite, quite tricky. The third point I always feel when it comes to mortality is most patients who benefit from any interventions are probably going to be in the medium sick category. If you are not very sick at all, you're more likely to anyway survive whatever I do to you. If you are very sick, you're more often than not not going to survive irrespective of what I do. And whatever intervention I do is going to probably change my mortality only in the middle group. And that kind of makes mortality a very, very ⁓ tricky outcome in my mind, especially for a simple therapy like an oxygen delivery device. So I feel mortality may not have been the ⁓ right outcome, primary outcome that I would have chosen if I had done the trial and we should have looked at other outcomes. Having said that, what did you think about the mortality data here? Do you think? ⁓ They are powered enough to do that. What are your thoughts about the mortality as an outcome here? Bharath Kumar T V: Yeah, yeah. Now I think that sort of lends itself to a nice discussion around how investigators come up with sample size calculations. And again, typically we look at four things, the type one error rate or the alpha, which is generally kept at 5%. That is the chance of declaring that you have a positive result when actually there is no difference, or finding a difference when there is no difference. And then we have the type two error rate, which is the chance that you don't find a difference, even though some difference actually existed. And we generally take this as 10 % or 20%. So then it gives you a power of 90 % or 80%. So these are relatively fixed, and there isn't much that investigators do with this. So it really brings down into two other things. One is the control group event rate. So what is the mortality that they expect in the standard oxygen up? So in this case, they are saying it is 18%, but that's, as you said, based on older data from their previous trial perhaps. And then the last thing that they have to postulate is what do they think is the expected benefit from HFNC, right? So in this case, they are saying it's 6%. So they're trying to get an absolute risk reduction of 6%. Now, if you look at historically at critical care trials, most trials, when they show a difference, they're showing a difference of less than 10%. Actually, for the vast majority, maybe even less than 5%. There are actually some nice papers out there that talk about this. sort of an effect maybe around two and a half percent. So whether six percent is a very optimistic difference ⁓ is something that we need to consider. So these are the factors that have gone into their sample size calculation. ⁓ But I think that's, Ramesh Venkataraman: I'm going to pass you there and say people have tried to compare and contrast Florali with the SOHO trial. I think sample size is going to hurt you both ways, right? It's very interesting. In Florali, they had only 300 odd patients in three groups and showed a mortality difference. Here you have, know, 1,000 patients in two groups and you don't see a mortality difference. And maybe, Bharath, it's worthwhile to elaborate on why small numbers can Bharath Kumar T V: Yeah. Yeah, yeah, Yeah. Ramesh Venkataraman: get outcomes too. Bharath Kumar T V: I mean, usually when we say, you know, we talk about this, we say the trial was not power to detect a difference. But in this case, if you show a difference in a small trial, in a small three-arm trial, it could still be a very much spurious effect. Right. So you still bring in the chance of a type one error. So in, you know, we can't take that, that those results, you know, we can't just take it on the, on the face of it, right. The, the mortality reduction shown in the smaller trial for sure. I think that's, that's definitely a concern. ⁓ So, given all this, ⁓ what are your broad thoughts? What do you think were the trials strengths, limitations, any concerns that you had with the trial? Ramesh Venkataraman: No, before I go to the strengths Bharat, I thought it's worthwhile making the point about the second, the secondary outcome, the first secondary outcome, which is need for intubation. Right. So need for intubation was 42 % in the HFNC arm and 48 % in the standard oxygen arm. So they did get a 6 % absolute risk reduction, not in the mortality though, but they got it in the need for intubation. So to me, working in a low resource country, so if you're saving one intubation and the Bharath Kumar T V: Yeah, that's good. Yeah. Ramesh Venkataraman: incidents of multi-drug resistant secondary infections in an intubated patient. Maybe saving an intubation is still an equally important outcome as a clinician that I look at. But if you do, I understand it's a secondary outcome and we are not powered for it, but if you just do the math, we need to probably do HFNC in 17 patients to probably prevent one intubation. Then the question comes, it cost effective in a setting like an Indian ICU where I do HFNC for 17 patients to avoid intubation. Is that still worthwhile for me to do? That's a question open to the audience out there to decide whether in their ICU do they still think it is worthwhile to go ahead and do that. But overall, Bharat, I felt it's a very well done study because only six patients were excluded out of the 1,116 patients who were originally enrolled. I think the comparison and the criteria were reasonably fair. they did have a reasonably objective criteria for need for intubation. And the commonest concern we all have of HFNC is that we may delay intubation. In this trial, if you look at it, the median time to intubation was also not different, which tells you that the delay did not happen. And there was no big difference between ventilator-free days between the two groups. So that again tells me that maybe they were quite right on in terms of having an objective criteria, even though is a little bit of subjectivity in it. I think they have done a really good job in terms of trying to figure out how to intubate. So, I feel these are all the positives of the trial in my mind. Any other thoughts? No, you have Bharath. Bharath Kumar T V: Yeah. Now, just to sort of add on to that, think one other maybe important point to mention is, we have to always assess the totality of the evidence. And in the case of HFNC, ⁓ there is a sort of consistency now evolving in the evidence, right? That there is no difference in mortality, but it seems to be consistently decreasing the need for intubation. So think that message seems to be coming through all of the previous literature, but also from this trial. So that, think, is useful. Ramesh Venkataraman: Mm-hmm. Bharath Kumar T V: And as you mentioned, there are several potential pathways from the time you connect somebody to oxygen to mortality. So whether mortality is the correct endpoint is of course a question. Otherwise, I agree with you on the need in a country like ours, avoiding intubation is perhaps a worthwhile objective anyway, because then you have infections and downstream problems from that. I had a couple of concerns. One is the fact that this is an open-label trial despite protocolizing when somebody needs be intubate. Intubation is very much a clinical decision and that is subject to bias. So that I don't think you can mitigate 100%. The second point which I think you already sort of mentioned when you're talking about the methods is the NIV rescue rates are pretty low. Only 3 % patients receive NIV. I'm not sure how they manage clinicians to not go down the path of escalating an IV, especially in the standard oxygen, right? Because it's a big decision to integrate from standard oxygen, ⁓ sort of skipping an IV, which might be our part. Ramesh Venkataraman: But then the interesting thing Bharath I found is the crossover was more from the HFNC arm to the standard oxygen arm rather the standard oxygen arm to the HFNC arm. That is something again I found, considering that the rescue NIV rates were low, I would have thought more people from standard oxygen arm would have gotten converted to HFNC. But the other way happened and that is something I really cannot explain. found it. Bharath Kumar T V: I'm not sure now. Yeah. Yeah. Yeah, yeah, yeah, it's a bit counter-intuitive, I agree with you, yeah. I was wondering whether the justification they had was that in their previous trial, know, Florali, NAV did the worst. So maybe they thought it's not a good idea to escalate to NAV, although subsequent evidence has evolved. Yeah, so, and then there's a minor point, you know, around, was not clear to me entirely how many were COVID patients. I know they said 314 from the previous trial, but nobody after that. But then in the supplementary appendix, there's more COVID, I don't know if there were co-diagnoses, Ramesh Venkataraman: Mm-hmm. Bharath Kumar T V: as in they also have something and COVID is not clear to me. But that was just a minor point that I wanted to bring up. But otherwise, yeah, I agree with you on everything else. you know, maybe it's a good time to talk about the take home, Ramesh. What do you think is a take home? Ramesh Venkataraman: Absolutely Bharath, I completely echo what you need to say. We all fall for the recency bias. The most recent one is the right one. It's very easy for us to be tempted to say HFNC doesn't work anymore and move away from it. think clearly you hit the nail on the head when you said that evidence has to be looked in total. It's almost like the steroid story, I mean, back and forth in terms of mortality and then over a period of time, mortality benefit may not be there, but then you have vasopressor that you can be enough. So same thing with HFNC. If you put all the evidence in front of you, I think it's more and more clear that it may not impact mortality, but clearly can possibly decrease the rates of intubation. And that's something that we need to be... ⁓ aware of and probably use it. I think it doesn't take away HFNC as a tool, but I think it narrows the scope of HFNC. I need to be very clear in terms of why I am using HFNC and how long I'm going to use HFNC. My concern is that we need to be able to have consensus in terms of how to decide HFNC success or failure. Unlike NIV, right now I think HFNC Now people use various scores, the high score, other scores to decide whether HFNC is going to be successful or not. My worry is if you are going to use more HFNC to avoid intubation, it shouldn't be delaying intubation. So that's something that we need to be ⁓ careful about. Those two would be my biggest take home messages. One, narrow down your scope of HFNC, be realistic about expectations. And two, be very, very clear in your mind when you would. No, go ahead and intubate somebody on HFNC rather than prolonging them on it. And the third becomes the resources in a country like ours. Are you going to invest in so many HFNCs so that you can avoid intubations? Is one more expensive than the other? I don't know. We need a robust cost-effective analysis to actually look at how much we spend on HFNC to avoid one intubation, and is it still worth it for a country like ours? So those are just some thoughts I'm going to leave you with, Bharath Kumar T V: Yeah. No, I want to just say that during COVID, know that a lot of the public hospitals also invested in HFNC, right? Like lots of it, at least in the major cities. ⁓ But I'm not sure whether they also had the backup to make sure that devices work, right? You need a high sort of a flow of oxygen, need a source that is appropriate, right? So pipe to oxygen source and so on. But otherwise, yeah, I think great points, great trial and ⁓ my take home message is similar that continue to if you have the resources, if you the device, adopt if you can, but yeah, with caution, especially with making sure that you monitor them closely. So on that note, we will conclude this episode. Thank you once again for joining us. Ramesh Venkataraman: Absolutely. Just want to tell our viewers, we're going to get an expert to talk about all this again shortly after this. And hopefully, the audience can enjoy both episodes back to back. Thank you. Bharath Kumar T V: Thank you.