Effect of Noninvasive Respiratory Strategies on Intubation or Mortality Among Patients With Acute Hypoxemic Respiratory Failure and COVID-19

This transcript is auto generated and unedited.

>> All right, well, thank you very much, Rob, for the introduction. And thank you for the opportunity to present the Recovery Respiratory Support Trial this evening. It's great, as always, to have these sessions where we can present the findings, discuss the findings with a panel, and the wider audience out there. So, thank you again. And we're -- I'm presenting -- we're representing this evening on behalf of the entire Recovery Respiratory Support Investigator Team. So, yes, the Recovery Respiratory Support Trial was funded by the National Institute for Health Research, and prioritized by the Chief Medical Officers as one of the urgent public health studies during the COVID-19 pandemic. On behalf of the Trial Management Group that you can see listed here, we'd like to thank the members of the Trial Steering Committee that is chaired by Kathy Rowan, and the members of the Data Monitoring and Ethics Committee chaired by Duncan Young, for their oversight and support during the study conduct. And, further, to highlight the Trial Management Team at the Warwick Clinical Trials Unit, who have been responsible for the day-to-day management and conduct of the study, and also the Sponsors Office for their support. So, just moving on then to the background, where did the Recovery Respiratory Support Trial really come from. So, we know that COVID-19 causes severe lung inflammation. And that, in itself, then leads to acute respiratory failure. Around three-quarters of hospitalized patients will require additional supplemental oxygen during their stay. And around 10 percent of patients will require admission to the intensive care unit for invasive mechanical ventilation. Recent data suggests more than 38,000 patients have now been treated in intensive care with COVID. And the latest government data indicates that just nearly 900 patients are in intensive care at the moment with COVID disease. And that's around 20 percent of ICU bed capacity. And, so, whilst this was thinking about the need to protect patients who are experiencing acute lung failure and acute respiratory failure, the need to look at noninvasive respiratory support strategies to protect against the need for invasive mechanical ventilation. And these center around controlled oxygen therapy, continuous positive airway pressure and high flow nasal oxygen. And, indeed, we know that all three of these treatment approaches have been used at various stages during the pandemic. And the emphasis has really been on preserving and protecting critical care bed capacity, and avoiding the need for tracheal intubation and invasive mechanical ventilation in acutely deteriorating patients. And, especially, during the early stages of the pandemic, when the clinical cost of COVID was unknown, and therapeutics that have recently -- that have since been shown to be effective, were not yet known at this -- at that stage. So, the emphasis was really on trying to preserve the bed capacity, and trying to ensure that we were able to meet the requirements for the numbers of patients requiring ICU admission. And you will recall that there was significant concern around whether we would have sufficient resources, and the numbers of ventilators required, to meet that surge capacity in critically ill patients. And not just within the NHS system, but thinking about healthcare services around the world, especially where access to these resources isn't as normally available, as we may be familiar with. And we, also, are familiar with seeing scenes like this around the critical oxygen supplies around the world, especially in lower- and middle-income countries, where oxygen supplies were so constrained that this caused clearly societal and patient care impact. And, in the UK, the NHS was relatively protected in that regard. Although we weren't -- didn't -- we didn't -- were unaffected by these critical incidents. And, certainly, we saw these, and a number of other incidents, occur at various stages, requiring the support guidance from the Intensive Care Society around managing oxygen supplies during the pandemic. And, so, against this clinical backdrop of varying management of patients, we also saw this multiplicity in various national and international guidelines and recommendations, around how best to treat patients with acute respiratory failure secondary to COVID. Many of these guidelines were contradictory. There was no consistency in the recommendations that were coming out of them. Many of them suggested CPAP. Others suggested high-flow nasal oxygen was the best approach. In many cases, the evidence that was informing these guidelines and recommendations was minimal in quantity, and indirect and low quality. And, so, coupled that with these varying guidelines, and also clinical empiricism and concerns around the various advantages and disadvantages around using different noninvasive respiratory support strategies, the requirement for a trial such as Recovery Respiratory Support was clearly evident. And, so, the main aim of the Recovery Respiratory Support Trial was to determine if CPAP or high-flow nasal oxygen was clinically effective, compared to conventional oxygen therapy in hospitalized patients with confirmed or suspected COVID-19 acute respiratory failure. How did we do the trial? So Recovery Respiratory Support is an open-label, adaptive randomized control trial, with three arms comparing the effectiveness of CPAP or high-flow nasal oxygen in those acutely unwell patients hospitalized with COVID-19 acute respiratory failure. The adaptive design of the trial meant that we were able to stop the trial early, if one or either of the treatment arms showed effectiveness, compared with conventional oxygen therapy. And our target sample size was just over 4,000 patients, aiming to recruit patients from more than 60 sites across the UK, over a 12-month period of recruitment. Our inclusion criteria were adults aged 18 years and above, who were admitted to hospital with suspected or proven COVID disease, who experienced acute respiratory failure that was characterized by oxygen requirements of at least 40 percent and above, and with saturations of 94 percent and below. These patients were required to be suitable for tracheal intubation, if treatment escalation was required. Our exclusion criteria was the immediate need for invasive mechanical ventilation within one hour, known or clinically apparent pregnancy, a contraindication to CPAP or high-flow nasal oxygen or clinical judgment, and where equipment for both CPAP and high-flow nasal oxygen were not available at the site at that time. And that features in with our randomization process, which was an internet-based system with allocation concealment. We knew that sites may either permanently or temporarily have access to only one or either of the intervention, CPAP or high-flow nasal oxygen, at any point during the pandemic. And, so, we factored that into the randomization approach. If sites had both devices available, they were -- patients were randomized to receive either CPAP or high-flow, or conventional oxygen therapy, on a one-to-one-to-one basis. And if either CPAP or high-flow only were available, then patients were randomized to either receive one of those interventions or conventional oxygen therapy, on a one-to-one basis. We didn't allow randomization between CPAP and high-flow only. And randomization was stratified by site, sex and age. In terms of the treatment arms, CPAP and high-flow were commenced as soon as possible, and conventional oxygen therapy continued delivered via face market -- face mask or nasal cannulae. But, importantly, across all participants in all groups, local policies influenced the decision-making around the choice of the device, the setup, the titration and discontinuation of treatments. We defined crossover as receiving CPAP or high-flow for more than six hours, when the patient was not randomized to receive that intervention, unless for the purpose of clinical stabilization, bridge to tracheal intubation or palliative care. An escalation will -- to tracheal intubation was performed when clinically indicated. And this was based on the judgment of the treating clinician. So, we didn't protocolized any of the treatment arms. And the treatment escalation was based on the judgment of the clinicians. Our primary outcome is a composite of tracheal intubation or death within 30 days. And tracheal intubation is a marker of requirement for invasive mechanical ventilation, which we know typically delivered in highest resource intensive care settings. Our secondary outcomes were the incidence of tracheal intubation and mortality at 30 days, so the two components of the composite primary, the time to tracheal intubation, the duration of invasive mechanical ventilation, time to death and mortality within critical care and hospital, the incidence of intensive care unit admission, and the length of stay within critical care in hospital. Our sample size was based on the data around conventional oxygen therapy rates at the time. We assumed a conservative incidence of 15 percent for the composite primary outcome, with a two-sided five percent significance level, 90 percent power. And that resulted in a 3,000-participant sample size initially, with 1,000 participants per arm, to detect a reduction of five percent. And we inflated that sample size to 4,002 due to the uncertainties underpinning those assumptions. And just a final slide around the methods around the speed and scale of the setup of the trial. Timelines like this will be very familiar to those of us who have seen the recent COVID trials, in terms of being able to instigate these studies at scale and speed. And you can see here that within the first 10 days from protocol to first patient recruitment, and that really speaks to the collaboration between the trial teams, the trials unit, and the research teams at all of the sites taking part. And, so, I'll now hand over to my colleague, Keith, to deliver the rest of the presentation.

>> Great, Keith, if you can open your slides, and get into SlideShare?

>> So, can you see my slides okay?

>> Yes, we can keep, Keith. Thank you very much.

>> So, thank you very much for that fantastic overview of the methods. I'm going to -- I have the privilege now we've been able to discuss the results. And I'm going to start off just by summarizing. The recruitment happened over a 13-month period. We were able to recruit 1,272 participants. We opened 75 hospitals. And, of those, 48 hospitals recruited in both the UK and Jersey. On the map, or on the side of the screen, you can see dots indicating each of our sites, but also the size of the dots indicating the amount of recruitment in each area. And what we can really see here is that -- very much that the recruitment that we saw matched the burden of disease were very much hotspots of recruitment happening in the northwest and the northeast, Birmingham, London and South Wales, and down towards Portsmouth as well. So, you will remember from Bronwyn's presentation that we aimed initially to recruit 4,002 participants. We stopped recruitment after 1,272 participants. And there were a few reasons for that. First of all, we'd reached the end of our 12-month funded recruitment period. But perhaps more importantly than that is this graph here, which shows the number of hospitalized COVID-19 patients within the UK, and also our recruitment, mapped alongside that as the dark gray bars. And what we can see is, when we reached the end of April and the beginning of May 2021, we really had reached an idea of hospitalized cases within the UK. And that, combined with the end of our funded recruitment period, and the desire to share the data that we had collected, and share that information with the international community, really drove our decision to stop recruitment at that time. In terms of enrollment and outcomes, we actually randomized 1,277 participants. Unsurprisingly, given the context in which this trial was conducted, that there were five patients who were accidentally randomized twice in error. So, we ended up with 1,272 participants included in our analysis. And the split of those was between CPAP, high-flow nasal oxygen and conventional oxygen therapy with the numbers shown on this slide. We have minimal loss to follow up, such that we have primary outcome data for 99 percent of our participants. In the Method Section of the presentation, we heard how we developed a randomization system that could account for the fact that sites may be able to offer all three interventions, or they may only be able to offer either CPAP or high-flow nasal oxygen. In terms of how that affects what happens with our patient flow, we -- it was clear to us that we wouldn't -- we weren't able to compare a patient who never had the opportunity to receive CPAP, because the site only had high-flow nasal oxygen. So, on that basis, we had a comparison of high-flow nasal oxygen that included all the high-flow nasal oxygen patients, with a group of conventional oxygen therapy patients, who had the opportunity to be randomized to high-flow nasal oxygen. Another comparison that compared all the CPAP patients with the group of conventional oxygen therapy patients, who had the opportunity to be randomized to CPAP. And on the side of the screen here, you can see the numbers of participants in each of those comparisons. Crossover is inevitably a challenge in this sort of trial, where there -- and we've seen it in many of the previous trials that have been done around noninvasive respiratory strategies. What we have here is a summary of the crossover that we observed within our trial. So, both CPAP and high-flow nasal oxygen had crossover rates of between 10 and 15 percent. With conventional oxygen therapy, experiencing a higher crossover rate, with 23.6 percent of patients either receiving high-flow nasal oxygen or CPAP, or a mix of the two. Our groups were well matched at baseline. This slide summarizes the overall population. Sixty-six percent of our participants were male. Ethnicity, 65 percent were white, 19 percent were Asian. Our mean age was 57 years. And our mean median time from COVID-19 symptoms to randomization was nine days. We highlighted earlier that the population that we recruited to this trial were patients that were considered suitable for escalation to tracheal intubation, if that was required. It's perhaps unsurprising then that we see a group of participants that are, when we look at their clinical frailty level, very much in the very fit, well or managing well categories. And that really does reflect the type of patients that might be considered for tracheal intubation. Thinking about the physiology of those patients that entered our trial, at the point of entry to the trial, at the point of randomization, the mean fraction of inspired oxygen was 60 -- was [inaudible] .61. Saturations were 93 percent. And 56 percent of patients were in a prone position, or had been -- or had been in that prone position at some point. The setup, as was highlighted, was left to clinical discretion. For CPAP, the mean peak was 9.5 centimeters of water. And for high-flow nasal oxygen, the mean flow was 50.8 liters per minute. And a reminder as to our primary outcome, which was our composite outcome of tracheal intubation or death within 30 days. So, on this slide, we summarize the primary outcome for both CPAP and conventional oxygen therapy, and high-flow nasal oxygen and conventional oxygen therapy. And what we can see is that the composite outcome occurred in 36 percent of patients in the CPAP group, 44 percent of patients in the conventional oxygen therapy group. And that gave an odds ratio of .72, with a 95 percent confidence interval that didn't transect one. So, we ended up with a significant benefit with the use of CPAP. When we compare high-flow nasal oxygen, we see the percent in conventional oxygen therapy, we see that the proportion of patients in each group that have the composite outcome was very similar. And we ended up with an odds ratio of .97, at a confidence interval that transected one. Which led to the conclusion that there was no benefit to the use of high-flow nasal oxygen in this group of patients. When we look at the breakdown for the primary outcome of mortality and incidence of tracheal intubation, these are the proportion -- or the percentages in each group. And I'm going to flip now to the odds ratios, because that really does help to -- to help us to understand what's going on here. In the -- in this CPAP versus the conventional oxygen therapy group, we can see here that the benefit really came from a reduction of tracheal intubation, rather than a reduction in mortality. So, that was the key driving force behind the benefits seeing that come as an outcome. In our comparison of high-flow nasal oxygen to conventional oxygen therapy, we can see that the odds ratios is again -- are very similar to what we observed with the composite outcome, that there was no benefit observed in either of those components. I'm going to move on now to the secondary outcomes. On this slide, we summarized the comparisons for CPAP versus [inaudible] oxygen therapy. We see here that with CPAP, we saw reduction in the percentage of patients requiring admission to critical care. And we, also, see that in the conventional oxygen therapy group, the time to intubation was shorter than it was in the CPAP group, in those that required tracheal intubation. And those that required -- and those in conventional oxygen therapy group had treatment for a median of one days before they received tracheal intubation. The duration of invasive ventilation critical care length -- stay in hospital length of stay, there was no difference between the groups. When we looked at the comparison of high-flow nasal oxygen and conventional oxygen therapy, we see here that across the outcomes are admission to critical care, time to tracheal intubation, duration of invasive ventilation, critical care length of stay, and hospital length of stay, there was no difference between those groups. One of the concerns with CPAP, in particular, was potentially around the impact on the patient, and some of the adverse events that might occur. When we looked at the safety reporting, for each of our treatment arms, we saw that adverse events and serious adverse events were more common in the CPAP arm than the other two groups. Thinking about some specific adverse events, we've picked out these four here. But, certainly, when we look at intolerance to the therapy, or the interface, a greater number of patients in the CPAP arm experiences adverse event. Hemodynamic instability was also numerically more common in the CPAP group. We've, also, summarized the incidence of pneumothorax and pneumothoraces across the three groups on the right-hand side of the slide. So, I think it's important here to recognize the limitations to our trial. But, first of all, we've had the great fortune here to be able to share our results at a very early stage. What this does mean is we've not had the experience of a very helpful peer review process. And we'll be going through that very shortly. We've discussed already that we didn't achieve our sample size, and the reasons why we chose to stop recruitment at the stage that we did. We -- obviously, in the context of this kind of trial, it's impossible to blind either the treating clinician or the participant. It's worth highlighting here that the -- with CPAP, the effect was driven by tracheal intubation. And that can be a subjective decision, with the variability both between patients and clinicians. However, we didn't see that same effect in the high-flow nasal oxygen group. But it is an important limitation to recognize. Finally, it's worth highlighting that these patients were assessed as being suitable for invasive mechanical ventilation. We don't know, and our results don't tell us, to what extent these results might be generalizable to the important population of people who are not considered suitable for invasive mechanical ventilation. And there may be a need for further research in that group of patients. So, to conclude, what we found was in hospitalized adults with acute respiratory failure due to COVID-19, CPAP compared with conventional oxygen therapy reduced the composite outcome of tracheal intubation or death in first days of randomization. In contrast, high-flow nasal oxygen compared with conventional oxygen therapy did not reduce the composite outcome of tracheal intubation or death in first days of randomization. So, I'm going to finish on this slide here. And this really is to reflect the fact that trials are extremely difficult to run in a normal context. What we saw with this trial was the coming together of a huge community. First of all, we'd like to acknowledge and thank the patients and families that supported the trial, to the doctors, nurses, Allied Health Professions Research teams across all the participating hospitals, who really did a phenomenal job in supporting the trial, and being able to offer participant -- offering -- offer patients the opportunity to take part. Trial delivery was supported by Clinical Research Network, both [inaudible] and NIHR in Northern Ireland. And we were, also, very grateful to Health Data Research UK, Office of National Statistics, and, of course, ICNARC for their support with data linkage and enriching our analysis. So, with that, I'd like to thank you all for your attention. And I'll hand back to Rob.

>> Keith, thanks very much. We will cross to London now. And Manu [phonetic] will deliver the editorial. Keith, if you could just close your screenshare please? And, Manu, over to you. Thank you.

>> Well, thank you. Can you see my slides?

>> Yes, they look great. Thank you.

>> Thanks, Rob. First, you know, thank you so much for the invitation, and the opportunity to discuss this trial. And a big congratulations to the trial teams for the incredible work, and the patients and relatives for contributing to this trial. Before I kind of start discussing the trial, I just want to acknowledge my funder, NIHR. I have an NIHR Clinician Scientist award. And my other research is funded through the various other grants as are listed there. And I want to highlight that these are my views, not necessarily those of the NHS, the NIHR or the Department of Health and Social Care. I do not have any conflicts of interest directly relevant to the talk, other than the fact that I know most of the investigators very well. So, what is the kind of clinical question or uncertainty that we are trying to address in the context of a pandemic? COVID-19, as Bronwyn highlighted, leads on to acute hypoxic respiratory failure. Then we have two separate questions really. First question is, "Should we go ahead and intubate early?" And a coherent argument has been made that we need not. And, therefore, what sort of noninvasive respiratory support should we provide? And which would be most efficacious? And the second point perhaps, not picked up until the discussion we have had thus far, is you have a strained healthcare system during a pandemic that is reaching capacity. And, therefore, could we safely spare some of the patient's either invasive mechanical ventilation or, in fact, shorten their duration of invasive mechanical ventilation, so that the capacity looks like it's bigger than what it is? So, in that context, when we look at recovery [inaudible] trial, the first thing that I would like to kind of say is, wow, amazing job for all those involved with the trial, in terms of the design to delivery within a short space of time. It is the largest trial of noninvasive respiratory support in the -- in COVID-19 patients, randomizing nearly 1,300 patients across 48 hospitals in the UK. A second important point here is, because of the kind of sophisticated data linkage and planning, the investigators were able to get 99 percent of the outcome data. And they show that there is a clinically relevant treatment effect for CPAP. They report an odds ratio that is adjusted, I'll explain the adjusted component later in the editorial, that you have an odds ratio of roughly 0.67, with a confidence interval of .48 to 0.98. And the graphs on the right highlight the CPAP in blue, compared to the usual care in gray. And the high-flow nasal oxygen and usual care again in blue and gray on the right side of the graph. As highlighted already, the trial stopped before reaching target sample size. And there were three interims conducted as part of the analysis with some [inaudible]. And, again, the point about crossovers has already been highlighted. So, with that sort of summary, I just want to kind of take the next seven to 10 minutes to talk about five specific points really, which are there is -- the -- you know, the trial is happening alongside a pandemic. The pandemic's phase is changing in terms of the clinical care. And the overall outcomes of patients over the duration of a year has changed. [inaudible] sophisticated covariate adjustment. It's worth kind of dwelling a minute or two on that. What are the implications of composite primary outcome, not generally for this trial, but outside of the trial -- outside of this trial, for this sort of intervention? It is inevitable that we will choose a composite of primary outcome. What is the implication of crossover? Which I am sure will be much discussed once the trial is published. And a bit about the clinical implications of this trial, starting with a very preliminary assessment of the risk of bias. So, what do I mean by changing clinical care? [inaudible] of hospitalized patients contributing to the [inaudible] consortium, and led by Dr. Anne Marie [inaudible], published -- published in Lancet Respiratory Medicine? What you see here is the data from Week 11, which is March 9th, to Week 31, which is 2nd of August, 2020. As you see here, over the course of the end of that window of 20 weeks or so, you get a peak, followed by a tail. And the mortality -- thankfully, inpatient hospital mortality drops down. And as you can see in the graph, towards July and August, the confidence interval is wide. Which, essentially, means that we are kind of reaching the tail end of that hospitalization episode. And, therefore, the mortality events are lower. The other important thing from this graph is the age distribution. Most of the admissions at this stage of the pandemic are kind of more than 60 years of age. When you look at the oxygen therapy, which are the proportions reported on the left-hand side of your screen, by admission week, from March through to August, you can see that the proportion of patients receiving supplementary oxygen gradually goes down. And the graph on the right compares the in-hospital mortality between the two populations, those who got supplemental oxygen and those who didn't. And the target population for recovery [inaudible] trial is the one that is in red. The mortality of that population is gradually going down. So, this is the target population that is likely to kind of get this intervention. And then you are thinking about the outcome, which happens in the intensive care setting. So, we'll look at the similar data for the ICU and mechanical ventilation. This is a paper from ICNARC, led by James Doidge, one of the statisticians there, and Kathy Rowan is the senior author. What you see here is we roughly grouped them into three kind of stages of the pandemic. And the -- what we observe is that, at the time of first 24 hours in the ICU, the [inaudible] ratio, just less than 200, the proportion of patients having that doesn't change. What changes, though, is the proportion of patients receiving invasive mechanical ventilation. So, when people read this trial, you know, this is the UK kind of clinical care changing, you know, across the period the trial is running. So, a very difficult trial to run. Again, a big congratulations to the -- to the trialists to have pulled this off. The second point I want to make is the covariate adjusted hazard ratio, or the odds ratio, that the authors report. So, the basic statistical premise here is that there are some baseline variables, or covariates, that are associated with the primary outcome here, i.e., need for intubation and mechanical ventilation. And, therefore, if you account for such covariates using statistical modeling, you can improve the efficiency of the trial. And you, also, avoid the bias from a chance covariant imbalance. And the authors carefully chose their covariates. Because, if you look at things that influence outcome, almost all of them are listed as pre-specified covariance. Which are age, as you get older, your risk of death goes up. If you are male, your risk of death is higher. If you have morbid obesity, your risk of death is higher. Ethnicity has an effect. Increase in [inaudible] has an effect. The target [inaudible] for the population here was .4. It was eventually .6 on average in this -- when they enrolled in the trial. Greater the respiratory rate, greater the risk of deterior -- respiratory deterioration. And the other added complexity that they had to account for is the fact that, around July 2020, a recovery trial highlighted that dexamethasone is effective. And the WHO Clinical Practice Guidelines and the prospective meta-analysis were published around the same time highlighting the role of corticosteroids. So, before and after corticosteroids, you are dealing with a slightly different kind of study cohort. And then come January 2021, we had IL-6 antagonists showing benefit in the prospective meta-analysis led by the WHO and the Recovery and REMAP-CAP Trial. So, the authors have accounted for a number of key covariates that could influence outcome. And, again, very well thought through, in terms of the analytic plan, to get this a priority specified. Then we come on to the issue of composite primary outcome. So, just a reminder of the composite primary outcome, if you haven't been reminded enough by the previous speakers. There is tracheal intubation or death within 30 days. So, I think this is an important point. And it is the correct primary outcome, in my mind, because the key for a composite primary outcome is that the -- it has to be a representative of the progression of the overall disease process. As a clinician, when I look at somebody who is progressing in their [inaudible] require tracheal intubation, it is a valid -- it's got face validity. And it also adds a few other benefits in the context of this sort of clinical trial. It increases statistical precision. And, therefore, the efficiency of the trial is improved. It, also, enables investigators to avoid some arbitrary choices around trial design, because you know fully well that, if you need to intubate somebody, their respiratory failure has progressed. And progressive respiratory failure means your risk of death has gone up. But it is not without challenge, I think, as I highlight in the figure on the right. So, the figure on the right kind of looks at the composite primary outcome distribution between the intubation and the 30-day mortality. And I think one of the important things to think about is that the constant -- if the constituents of the prior -- of the primary composite outcome do not move in line with each other, we have a problem. But it does seem that they are. The second issue to think about is each element of the composite primary outcome should have a similar treatment effect. And that may not be happening in this intervention that we are seeing. And the third, and perhaps the most important driver here, which is -- which could contribute to some of the inferences that we are making, is that intubation is a clinically driven outcome. And, therefore, that could influence the way we kind of deal with the primary outcome here. And the next point I want to make about is the crossover. And I think both the previous speakers highlighted the point about crossover here. In the CPAP, the standard care arm, roughly one in six patients crossed over. In the high-flow nasal oxygen, the standard care arm, you had roughly 12 percent crossover. And those who were randomized to standard oxygen, there was a crossover of around 25 percent. So, the key thing to think about, when we discuss this topic of crossover within a randomized control trial of an intervention like this, is think about the direction of bias and treatment effect. So, if you follow the argument, if you are randomized to a CPAP arm, and you cross over to the standard oxygen arm, what you are essentially doing -- and the trial is showing benefit with that intervention, which is what the authors have highlighted. What you would do with that crossover is to bias the results towards no, as in CPAP benefit is taken away when you cross over to the standard arm. And I'm not going to go through the other parts of the debate. You can -- you can make a similar argument for the standard oxygen to a CPAP on crossover, where you're going to bias the results towards no. So, despite this happening, have -- you know, seeing a signal for CPAP reassures me that that's probably likely to be a valid signal for us to kind of consider kind of in the clinical management. And the second issue to kind of think about in this crossover is the fact that it dilutes the treatment effect. That is what I meant by biasing towards no. And this is well established in the statistical literature. This is a -- this is a rather old kind of paper that we were taught, when I was kind of learning epidemiology at the London School, that actually if you -- if you -- if you have a trial with a fixed sample size, and you crossover, and you do an intention to treat analysis, what you're essentially doing is diluting the treatment effect of the intervention. And, here, despite the crossover of one in six, I think the benefit of CPAP is retained. So, again, I'm more inclined to believe the result rather than not. So, the last couple of minutes about the question that we all face, when we read a clinical trial that's coming along, that answers a valid question, "Should this trial change clinical care, or my clinical practice should be -- should it be modified?" And, there, it's a slightly more nuanced and sophisticated argument that you -- that we need to follow. So, the first point, I've used a risk of bias tool by Jonathan Sterne and colleagues published in BMJ a few years ago. So, that there are a few elements of the bias that we are thinking about. The bias from the randomization process. A green dot means that the bias is low. And given that the randomization, there was allocation concealment, and the randomization could happen one-to-one-to-one, the reason why I haven't given a dot -- green dot here is because you -- as Bronwyn highlighted in the methods speak -- methods lecture that actually there are patients who could be randomized either to high-flow nasal cannula or to conventional oxygen therapy, as two arms, when they -- when you did not have CPAP. And I think that's an important kind of thing to keep in mind. But I don't think that that is significant enough to kind of change the inferences that we are making about the trial. The second element of the bias is due to the deviations from intended interventions. And, again, I haven't given it a read. I've given it kind of -- sat on the fence to give it a slightly closer to read, partly because of the crossover. And the bias due to missing outcome data is not an issue for this trial, as you have 99 percent of the outcome. And the bias due to measurement in outcome is not a major issue, unless you are someone who believes that the blinding -- lack of blinding is a problem. But then that is the problem in any noninvasive ventilation trial. That is a problem in any prone position trial, and so on and so forth. So, you can't use that to make an inference that this isn't -- this is not a valid outcome. And there is no selective reporting of results, that I can see, from the statistical analysis plan that I have looked at. So, in summary, my risk of bias assessment here is that there are some concerns in at least one domain of the result for the primary outcome. I haven't done it for all the other outcomes. But I don't think that it is of any concern, in terms of the value of the information that we have been provided. So, the next point you would think about as a clinician is what else is out there in terms of clinical trials, and should I look at them, and see whether that is -- that gives a different answer. There are two clinical trials. First is the Greico Clinical Trial published in JAMA, comparing helmet NIV versus high-flow nasal cannula. And their primary outcome was the number of days free of respiratory support. And they did not observe any difference. And the second is a 22-patient trial comparing high-flow nasal cannula with conventional oxygen therapy. And I searched their published manuscript. I could not find the primary outcome. So, we're not looking at a multitude of high-quality trials that's out there which -- with sample sizes reaching 1,200 or 1,300. Although size is not always an important thing here, in the context of the evidence that we are looking at, it is relevant. And I just want to kind of talk one specific point about the [inaudible], the gray [inaudible] randomized control trial. That comparison as a highlighted was helmet NIV versus high-flow nasal cannula. One of their secondary outcomes was the hazard ratio for endotracheal intubation. What you see here is the helmet NIV, although a different intervention seemed to have a lower cumulative incidence compared to the high-flow nasal oxygen, the difference probably here, as Bronwyn highlighted, the decision to intubate was protocolized in the HENIVOT trial, whereas it was not in the recovery. [inaudible] that may have contributed to some difference. But I don't think that the -- that is significant enough for us to, again, say that the results aren't valid. The last point I want to make is that you don't look at evidence just on its own. You look at evidence from other kinds of publications. And this is a systematic review and meta-analysis of the noninvasive oxidation strategies for in [inaudible] mortality in acute [inaudible] in respiratory failure. And the authors did a systematic review, and identified 25 randomized controlled trials involving roughly 3,800 participants. And they did work with what we would call as a network meta-analysis. And in network meta-analysis, what you are doing is comparing the usual care with other interventions that are published. And, in this case, they have a comparison between face mask noninvasive ventilation versus standard oxygen. There are around 14 trials that address this question. That's probably the biggest evidence base you've got. And there are other trials that contribute lesser data to the evidence. So, the thicker the line, and the bigger the circle, the larger the evidence base. And what the authors conclude here in this meta-analysis is that, if you use a face mask NIV, you reduce the risk of mortality. And this -- the certainty of evidence is low, but the strength of association -- sorry, but this direction of association for the other outcomes we just need for intubation is consistent with the primary outcome. So, to answer my last question, which I posed at the start of the last few slides, I think the recovery [inaudible] trial will make me change the way I look after patients. Especially as it would make one person avoid invasive mechanical ventilation for every 12 patients treated with CPAP, instead of standard oxygen therapy. Perhaps the -- like with any editorial, you would conclude by saying, actually, you know, there are situations where you don't want to say CPAP is better than high-flow nasal cannula. And this trial doesn't necessarily answer that element of the question, I think, given the -- given the sample size, and how the trial had to be curtailed. But the primary signal, I think, is valid. And I kind of leave you with that slide set. And thanks again, Rob, for the opportunity to discuss, and congratulations to the trialists.

>> Manu, thank you very much. That was a wonderful editorial. I'll now invite Danny McAuley and Gavin Perkins, the co-chief investigators, to reply to some of the points that Manu has made. Danny and Gavin, over to you. Maybe -- Danny, maybe you want to go first?

>> Thanks, Rob. I guess, first off, just, Manu, thank you for such a balanced review. We will include you as a reviewer on the paper in due course, whenever we submit it. So, I guess I want to pick up on maybe one or two points. And I think you touched on the concept of the composite made up of a subjective outcome, in terms of need for intubation in the setting of an unblinded study. And I think that does, you know, merit some discussion as a limitation. I guess one of the discussion points will likely be that perhaps people were more comfortable pushing on with CPAP, in the event that they knew what the patient was receiving, and they might have pre-existing biases about respiratory support. But you would imagine if that was the case, that that effect would be seen both with CPAP and high-flow. And I think that the fact there's a differential between these two forms of noninvasive support, to me, would suggest that there might be a genuine effect, as you sort of alluded to with CPAP. Because we don't see that same effect with high-flow. I really liked your highlighting of the crossover. I think that's really important. It's a challenge with all noninvasive studies. And I think, you know, we've got to overemphasize it, but I think it is worth highlighting that the crossover would null to no effect. And I think that's a really valuable point. And then I think my last comment, because we're pushed for time and I want to give Gavin the opportunity to pick up on things as well, is you rightly say that this wasn't protocolized in terms of when to intervene. And there was always a concern that people were commenced on CPAP or high-flow, and left for prolonged periods, and that might be associated with harm. And I guess it is important to highlight that our trial doesn't really get at that question specifically. But it is reassuring that the average time on CPAP was only four to five days. And there was a delay to intubation in those who eventually needed intubated in the CPAP group of two versus one day approximately. But that didn't translate into any evidence of harm, or particularly a signal for increased mortality, as you say. Although not significant, the mortality goes the right way. So, I guess there were my initial comments on your very thoughtful editorial. And, again, with that, I will hand over to Gavin to pick up.

>> Oh, thanks very much, Danny. And, again, just to echo those comments, with thanks to Manu for a carefully thought out and presented editorial, I'll probably just like to make two points in response. First is to pick up your observation in terms of the choice of the primary outcome. And, as you can imagine, this was something we spent quite a bit of time thinking about at the start of the trial. And I guess when one looked back at the evidence base, there was really a mixed picture as to the likely effects of CPAP. Whilst there were some observational studies suggesting potential benefit, I think we're also alert to some of the studies in a critical care setting, where the use of CPAP has delayed intubation. And, as a consequence, lead on to mortality. So, to separate out intubation, and not have a measure of mortality, you know, I think was going to be challenging. I actually think the way that the results have turned out help us to answer the question. I think you're right. The findings were driven by a reduction in the need for invasive ventilation. But the pattern that we saw was also in relation to whilst there was no significant difference in mortality, at least in absolute terms, we didn't see a signal for increased mortality associated with CPAP use. And I think my final point is -- goes back to probably your first point, which is changing clinical care. And I guess, you know, one of the strengths, and it is a real strength of this study, is it shows how the respiratory and critical care community and the National Health Service came together to deliver one of the largest randomized control trials in the middle of a pandemic. It engaged with patients and their families. And we're incredibly grateful to their help and support. But I, also, think it reflects on the great infrastructure that we've got in terms of ICNARC, because, we'll be honest, if it wasn't for ICNARC helping us with the -- you know, the data linkage, we wouldn't be here presenting to you now. We'd be several months away from being able to share this important data with the wider community. So, with that, I'll thank you once again. And hand back to Rob.

>> Gavin, thank you very much. I'm going to bring in my colleague now, Chris [inaudible]. Chris has questions from Twitter and the chat function. I think he's going to initially direct them to Danny and Gavin.

>> Thanks, Rob. So, first of all, the audience was very impressed by just how quickly this trial was set up and running. So, there's a lot of congratulations on how well that was done. And a few questions, first of all, just on this select -- the trial had quite simple selection criteria, [inaudible] less than 94 and 40 percent oxygen. And there's a few questions about [inaudible] can explain why this was selected. Because it appears that that would be a very large population of potential patients. And do all of these patients then potentially need to the CPAP? Or can the author suggest who would benefit the most from it?

>> So, I'm happy to take that first. And, again, I think it's a great question. I guess we were wanting to set up, you know, a study in the context of a pandemic that would be intervening early enough to potentially have some benefit, so that there was an opportunity to mitigate the need for invasive ventilation. If we left things much later, there'd be a concern that potentially that there wouldn't be a sufficient therapeutic window for CPAP to, you know, potentially reduce the need for invasive ventilation. I think a key message, though, that came out from the study that despite the fact that we set an FIA to a 40 percent as the inclusion criteria and stats of less than 94, the patient population that were recruited appear to have been a little sicker than that. I think the median FIA to was not .6 percent. And I think, you know, this shows that giving clinicians in the NHS a protocol, and allowing them to apply their clinical judgment in that setting, is what's delivered this trial result. So, I think I would take from this that 40 percent and a SATS of 94 percent is the minimum level. But add that alongside your clinical judgment.

>> And from having done the trial, can -- would you suggest should all of these patients get CPAP then? Or should you be a little bit more selective?

>> So, I don't know if Danny wants to take that one.

>> Yeah. I mean, I think it's a similar response to essentially what Gavin said. I mean, I don't think that we can get at that question specifically. We have our eligibility criteria. But, often, whenever you establish eligibility criteria, people tend to recruit a sicker population. And we often see that in many respiratory failure studies. So, I think the conclusion is whenever you see someone whose oxygen requirement is increasing above .4, and their SATS are not being maintained, you know, above 94, that's the time to start thinking about CPAP. And, you know, I guess if that is rapidly increasing at that point, then CPAP is probably appropriate. If things are very stable at that point, it might be reasonable to wait and watch. But, ultimately, I think the message is that whenever you're going to use a form of noninvasive respiratory support, this evidence would suggest that that should be CPAP.

>> Great. There were a few mentions of the different results of this trial in comparison to the Feralli [phonetic] trial. And [inaudible] question. Can you account for the difference that CPAP seemed to be superior in this trial, in comparison to high-flow, when in the Feralli trial may seem to be equivalent?

>> Yeah. I mean -- Gavin, I'm happy to take that one. Yeah. So, this has been asked -- I mean, I guess, you know, it's getting at what is the mechanism by which CPAP works in this disease, as opposed to, you know, previous studies which have reported, in particular, community acquired pneumonia. And that, you know, CPAP didn't have a benefit. And might actually have higher incidence of adverse effects. And I guess it gets at the, you know, potential pathophysiology. And it may be that there is more likely to be recruitable lung in the setting of COVID-19, which maybe we don't see in other situations of acute hypoxic respiratory failure. And, hence, why CPAP with a larger distending pressure might have better effects, compared to high-flow nasal option where, although it gives some level of [inaudible], it wouldn't be the same degree. So, that's a hypothesis that, you know, is in need of, obviously, potentially testing. But that would be one explanation that there's the opportunity to recruit more lung with CPAP and this condition that isn't seen in other conditions.

>> And just why we're met -- we've mentioned the Feralli trial. There's a question relating to why this trial compared CPAP against standard therapy, and high-flow against standard therapy. Was the Feralli trial compare -- had a three-way comparison? Can you expand on that for us?

>> So, I think I'm happy to take that one. And I think it was from a statistical basis, that the more comparisons between interventions that you're doing, such as the largest sample size that you need. So, we weren't -- we didn't set out a priority to compare CPAP with high-flow. We were wanting a head-to-head comparison with controlled oxygen therapy.

>> Okay. And there's just a couple of questions relating to the fact that there wasn't a -- there wasn't a lot of the CPAP or the high-flow protocolized. And that the indications for intubation weren't protocolized. Could that have had an effect on the outcome, given the changing management that we saw during COVID? So, we saw initially recommendations of early intubation, which seemed to change to maybe a preference for noninvasive.

>> So, I [inaudible] that one, if that's okay, Danny?

>> Yeah.

>> Yes. So, I think, again, one of the strengths of the trial was around the fact that randomization was taking place throughout the pandemic. So, we were having equal numbers of patients, randomized reaches with different arms, as time progressed. And, so, any temporal changes in treatment ought to have been -- remained consistent around the trial. In terms of whether we would have been better to, you know, protocolized the decision for intubation, and protocolized the use of the interventions, I guess, you know, in terms of trial design, there are two types. There are efficacy trials, and there are effectiveness trials. An efficacy trial would tell you if something works under the ideal conditions. And an effectiveness trial tells you about how it works in real world conditions. And I think the question that was important for us, for the NHS, and for patients, was how it worked in real world conditions.

>> And just to pick up on that. The concept that you could get a bunch of ICU doctors to agree what the criteria were for intubation would be pretty challenging as well. So, I think [inaudible] --

[ Inaudible Speakers ]

>> [inaudible] recommended criteria.

>> [inaudible] living.

>> I suppose then, just following on from that, in real world conditions, is there any sort of data or any feeling for where these interventions actually occurred? So, the length of ICU stay seem to be quite similar. So, I mean -- so, is there a feeling that maybe CPAP on the ward, when moved to ICU, would have been more effective? And, therefore, that it was really ICU care had the effect on the outcome? Whilst high-flow maybe could have stayed -- or stay on the ward for longer periods of time?

>> Yeah. I mean, that -- that's a good question. And, Gavin, I'm happy to take that one.

>> Yep.

>> So, it's a really important question to understanding where process of care was delivered. I mean, I think during the course of the trial, in terms of where it was delivered, the majority of the interventions were delivered on the ward. I mean, I think, as we all experienced, you know, that the indication for admission to intensive care was need for invasive ventilation. So, the majority of the interventions were delivered on the ward, typically, as the pandemic progressed in respiratory support units. And I guess, as Gavin alluded to, you know, this couldn't have been delivered without the collaboration of our respiratory physician colleagues. So, we do have that information. The majority of it was ward-based. There were some people who were removed for the intervention to ICU. But it was, you know, the minority rather than the majority. So, although that is a potential confounder, I think it's a less likely issue to explain the findings.

>> Okay, thank you. And there's a few other questions, if we have time, Rob?

>> Maybe just one more, Chris. I think we're just [inaudible] --

>> Okay. [inaudible], yeah, there's a very practical question. And Manu has already sort of discussed about it. But from a trial organizational point of view, how do you deal with a 25 percent crossover? Do you [inaudible] or -- how do you deal with that when you're interpreted in your results?

>> [inaudible] shall I take that one to start with, Danny? So, I think in terms of planning, as we went into the study, you know, we thought it was inevitable, as most previous trials of noninvasive ventilation strategies have included crossover, that it -- that it would occur. I think [inaudible] already answered on, and I think as Manu covered in his editorial, the effective crossover actually would be to potentially reduce the treatment effect as, in essence, the standard oxygen therapy, what we saw with the group that crossed over more frequently. So, the fact that despite, you know, moderate levels of crossover, you know, we nevertheless saw a signal of benefit for CPAP, where we didn't see a similar signal of benefit for high-flow.

>> Thank you.

>> Great, Chris, thank you very much for that. Some fantastic questions. Hopefully, I'll have time to get to some more later on. If not, then we will try and do a podcast, and mop up some of these questions that we haven't had a chance to answer tonight. What I'm going to do now is I'm going to move to our panelists, who have been patiently waiting. And I will ask each of them, in turn, just to give their initial impressions on the trial. And if they have any questions for the trialists to pose them as well. So, Leticia, you were first in this evening. So, I will hand over to you in Paris.

>> Thank you very much. It's an honor for me to be here. I want to wholeheartedly congratulate the Recovery Respiratory Support Team for doing this such a great job. It's a huge accomplishment to recruit all these patients, throughout the pandemic, your delivery and important unmet need to the intensive care community. So, congratulations for that. I have a few points that is really intriguing for me, from my perspective in Brazil from the pandemic. So, for example, in Brazil, I think we wouldn't be able to design a trial with a low-flow that goes straight away to intubation. So, I think that your crossover, 20 percent, from an operational perspective is something that is pretty good. And another comment, also bringing the perspective of low- and middle-income countries. When I look at your results, and I see that there was no [inaudible] on mortality, when the patient came from low -- the group of low-flow, whether had CPAP or low-flow, this is somewhat reassuring for settings that are short on resources and devices. So, I would like to hear your thoughts on that, on looking to the other side of the coin.

>> So, I don't know whether Keith or Bronwyn wanted to respond to that.

>> So, I'm -- I can respond. So, I guess, the first thing you raised was around the crossover. I guess that really reflects the commitment of the people throughout the recruiting hospitals, and their real dedication to helping us deliver a trial that answered this clinical question. I think your point around the challenges in other countries, with perhaps fewer resources, is an important one. I guess it -- you know, at the very least, our trial provides reassurance that, if you were to decide not to give high-flow nasal oxygen, and conserve your oxygen supply, and to continue to give conventional oxygen therapy, that would be a very reasonable treatment strategy, that wouldn't be disadvantaging the patient in any way.

>> Thank you very much. I have a bunch of other questions. But I'll go to Twitter later. And now just to leave the opportunity for the other panelists to talk. Congratulations again.

>> Thanks, Leticia. Janice, you've been waiting patiently as well. Your thoughts, please?

>> Thank you. And, likewise, like my colleague, Leticia, I would like to just congratulate the investigators. This is a wonderful study. And, also, just to acknowledge that the studies that have been done, and the trials that have been conducted, during this pandemic really show the great collaboration on a global level, in all the regions, of how it is possible to do good research during a pandemic, and to answer key questions. And I think that was really one of the biggest lessons we've learned from this pandemic, that we can find answers, and impact clinical care in an evidence-based way. So, I just want to thank the colleagues for doing this study. As already discussed, that the numbers that have been enrolled in this study really are large, compared to what we know. And this is in COVID, as opposed to the evidence we have come out before COVID. What we will do with this information, that has been now presented and shared with us, is really to do a little bit of what Manu do -- a little more of what Manu do, is to critically look at the totality of the evidence to see what does it show, how certain are we on these results. And from what you've shown, I think it is looking very promising for those for CPAP, in comparison to conventional oxygen. And, so, that is, can -- you know, will impact our clinical guidelines, and will impact and have impact on clinical practice, instead of something that is kind of now on our side to kind of take this evidence, put it within the context of what we already know, and how can we derive some clinical practice guidelines for that in a timely fashion to impact patient care. At the same time we look at resource utilization, and I think that's a little bit of what Leticia was getting at. And, so, now is a CPAP machine available in low-, middle-income countries, which is, you know, some of the major stakeholders that look to WHO for clinical guidance and support? Are the oxygen systems sufficient to run CPAP at these levels and settings? However, acknowledging, though, that if you reduce invasive tracheal intubation and basic mechanical ventilation, then actually you are likely conserving oxygen in general. But there are decision points there with the resource utilization, and oxygen access and availability that I think we do need to look for -- look -- look more into in the use of CPAP. And, also, we were looking in for high-flow nasal oxygen compared to conventional oxygen. But I think the results are very promising. And I'm happy to hear this. I think what I would ask the investigators are, "What do you think would be the next -- what do you think are the necessary studies to do to answer the questions, if it should be still answered, about the role of high-flow nasal oxygen? So, now that we see this benefit potential -- this benefit for CPAP compared to conventional oxygen therapy, you know, at a relatively early transition point, what do you think is necessary to look more closely at the question between high-flow nasal oxygen and CPAP? Or do you think that this is -- that answers that question? Over. Thank you again for the opportunity to be here.

>> Thanks. So, [inaudible] I can pick up a little bit on that question then thinking about what the next unanswered questions would be. We touched upon the fact that we didn't include patients who are for escalation of treatment. And that may be a population that we need to look at some of the other alternatives. And high-flow may still have a role with those patients. And, so, there are definitely cohorts that we pragmatically didn't include at this stage. And I think those patients who aren't -- the clinical judgment isn't that they are for escalation to intensive care admissions still require us to optimize their management. And that, in itself, would have resource implications, both for in terms of practical, but also patient care level as well.

>> Fantastical. Go to Steven now in just a moment. Can I remind our viewers, if you are on Twitter, to please use the hashtag Recovery-RS, just so we can keep track of your questions and comments. So, Steven, in Cambridge, you've heard quite a bit now tonight. What are your thoughts on what you've heard? And do you have any questions for the trialists?

>> Thanks, Rob. Yes, good evening from Cambridge. Not quite so exotic as Paris, Sao Paulo or San Francisco. But nice to talk to you, nonetheless. Three points from Cambridge this evening, and three points, I suppose from the Intensive Care Society, which I'm representing tonight as well. I mean, first and foremost, this is a huge success for UK Critical Care Research. Absolutely fantastic effort by everybody involved. The Intensive Care Society are delighted to have supported the trial from the beginning. And it's really fantastic to see, and be part of, the trial results presentation this evening. Danny and others will know I have had conversations while the trial was going on back and forth. And to see it [inaudible] in fruition and be part of the results is fantastic. So, yet another success for UK Critical Care Research. And a big mention as well for all the other trial participants, researchers, [inaudible] -- critical care, emergency medicine, respiratory medicine, the doctors, nurses, and the patients and families who were involved. So, well done to all involved. The second point, I think really relates to the choice of modality for acute respiratory failure, which is a really key question for clinicians who are facing COVID in this wave. And, you know, we never know what lies in front of us. But I think the key choice, that is the reassuring safety message that is coming, that CPAP appears to be safe to use, that it appears to be more effective than high-flow nasal oxygen, and more effective than conventional oxygen. So, that has significant implications for UK critical care units and our, you know, Intensive Care Society clinicians and Intensive Care members, because we know that there is a lot of discussion and anxiety about use -- about use of oxygen, and also about infection control practices regarding aerosol generating procedures. So, the question here -- the key message here is that, if you have a patient requiring oxygen deteriorating, who needs escalation of care, it is -- and you want to use a therapy which will generate aerosol, then the most effective aerosol generating procedure to use would be CPAP, as opposed to high-flow nasal oxygen. That will give you a treatment benefit and a mortality -- and a reduced reduction in tracheal intubation, and should bring benefits overall to the patient. And I think the third message really from us is the -- is the importance of the critical care or -- not the critical care, but the respiratory care that is being delivered outside intensive care units. The importance of having appropriate organization, infrastructure to deliver this sort of care outside critical care units. We've heard discussion, a lot of discussion, about acute respiratory care unit, about enhanced respiratory care, making sure that hospitals in the UK, and across the world, are set up to be able to deliver these sort of therapies safely with appropriate resources, appropriate staffing outside the critical care unit, so that bed capacity and intensive care can be preserved and maintained. But, also, that patients get the appropriate care that they need outside of intensive care units. So, three thoughts from me. And then a final question. The question relates to that old chestnut of CPAP, and the question of safety, harm and efficacy. Were the trialists surprised in any way at the level of harm that was fine to be associated with CPAP during the trial?

>> So, thanks, Steven. So, I think I might pick up on your AG, aerosol generating procedure comment. And then I might hand over to Gavin to pick up the question about safety and harm and efficacy. So, it's -- kind of almost might be accused of paying you for the aerosol generating procedure comment. It is very interesting, and has been the subject of much debate. And, in fact, it highlights the benefit of peer review. As we were setting up this, everybody said, "You must find out what happens in terms of aerosol generating with these interventions." And with the help of a really great bunch of people, [inaudible] led by Wendy Barclay, who's an expert virologist from Imperial, and Chris Green, who delivered a substudy from Birmingham, we basically looked at whether or not we could culture virus in the environment, according to the use of the intervention. And, essentially, it doesn't matter what intervention you're on, it's probably patient characteristics rather than the intervention. So, you know -- and that's in preprint at the minute as well, if you want to go and look at it. It's a relatively small study, but still quite convincing that, in fact, I think we probably need to be aware that aerosol generation is driven by the patient, more so than the actual intervention. So, I'll maybe hand over to Gavin to see if he wants to comment on your second question.

>> Thanks, Danny. And thanks, Steve, again, you know, for a great -- you know, great commentary. So, I think your question was whether we were surprised to see a stronger, you know, safety or harm signal in the CPAP group. And I guess, to an extent, if we look at the [inaudible] data that's looked at the frequency of pneumothorax and barotrauma, I guess we weren't that surprised to see that the incidence was greater. Because I think in that large observational study, there was almost a stepwise increment that controlled oxygen therapy, as we saw had the lowest rate of complications. And then CPAP, high-flow stepping up. And then the combination of all three with invasive mechanical ventilation, you know, with the group that sustained the most harm. And I guess, you know, in considering, you know, it in its totality, you know, I guess we come back to, you know, the primary outcome, or the two components of the primary outcome. Which is that, overall, there was a reduced rate for intubation with 12 patients -- or for every 12 patients being treated, [inaudible] requiring invasive ventilation. And, you know, whilst we saw no difference in the mortality, you know, figures, in absolute terms, you know, that the number, you know, wasn't higher in the CPAP group. So, I think, you know, there's a message there to be cautious, you know, that I think is part and parcel of delivering the intervention. And having knowledge and foresight of that I think we're better equipped clinicians.

>> Okay, [inaudible] to bring John in at this point. He has been listening for quite a while. John, can we have your thoughts, please?

>> Many thanks [inaudible]. First of all, just to echo the enormous congratulations to the group. It's brilliant seeing a trial of this set up so quickly, and, you know, within a short space of time achieving its objectives. There's so many points that this could have gone wrong. And it really is brilliant that you had pulled it off. And I'm -- you know, I say that because it gives encouragement to the adoption of these designs, in less sort of febrile times, that they can actually be deployed, and generate what looks like very, very good evidence that it'll be difficult to argue against. This -- you know, one problem about starting very quickly is pushing out to make some very quick assumptions. And, you know, not being critical, but you got -- you got the event rate badly wrong. You know, it -- I think you were in for a 15 percent control rate. And it turned out to be about three times that. And this is actually a feature of these types of trials, that the assumptions often let you down. And what was really, really good in this trial is that you had a very sensible set of sequential monitoring plans that were very flexible. I think you were just kind of spending the [inaudible] as they call it in a linear fashion. So, it [inaudible] really matter. You actually ended up observing roughly the number of events that you expected to observe in 4,000 patients. You got about 500 events. And, therefore, you didn't find yourself kind of beached on the altar of this kind of statistical complexity. Which was really, really good. I think probably got to be a little bit more careful about this crossover business. It is not -- it is not as straightforward as some of the speakers have been making out. You know, the problem is that when we talk about crossover and diluting the treatment effect, we're kind of implicitly assuming that the people that crossover get the benefit of the treatment. And that probably, in this case, is not the case. They're getting crossed over because they're deteriorating. And if you -- if you then -- you know, it might be that the people that crossed over to the -- to the more invasive oxygenations didn't get any benefit at all, if they shortly afterwards needed to be intubated or died. And you've got to be very careful if you move away from the ITT, and start doing more sophisticated statistical analysis such as causal modeling, or as the -- as your protocol, your SAP suggests some sort of inverse probability weighting type thing. Because you can quite quickly get the wrong answer in the other direction by doing that, and take something that actually works and persuade yourself that it doesn't work. Composite outcomes, I think, you know, they're difficult, and you do have to be careful with them. But I think the composite here is reasonably coherent. It looks like most of the people that died were intubated before they would die, just looking at the way the numbers fall. And, so, yeah, I'm just really -- I'm fascinated that your system analysis plan said you were going to do these additional analyses. And, also, possibly some Bayesian analyses. So, I think -- you know, when you get it published, I think there will be a lot of interest in getting as much out of the data as possible. And, certainly, I'm -- as a methodologist, I'm fascinated seeing it. But just the most important thing, fantastically well done. And you're superstars, you're heroes.

>> John, thanks very much. I'm going to nip it here before Danny replies. Just a quick question. Manu has mentioned this concept of alpha spend. Can you very briefly just explain to our listeners and viewers what that is. Many of the clinicians may not be familiar with that.

>> I think it's probably beyond my capabilities of explaining it briefly. It's one of the -- one of the more unpleasant aspects of designing trials in a frequentist way. It's the -- it's the idea that if you -- if you look at the data repeatedly, you are increasing the chance that you're going to come to the conclusion that treatment works. When in truth, it doesn't. That's the so-called type one error, false positive type error. And the idea is that you -- if you specify how many times you're going to look at the data, and, in this case, there were 11 interim analyses specified, as it happens that was kind of abandoned after about three or four, and they just left it to the [inaudible] to decide, because it -- I think it was proving to be too difficult to keep sys control. But the idea is that they will have done three interim analyses. And then the p values that you'll see in the report when it's published, they're using a statistical method, which actually adjusts those p values for the number of times they've [inaudible]. So, crudely, instead of having a nominal level -- an alpha level of five percent, the actual operational alpha level will be something more like I would imagine about 3.8/3.9, because they were spending the alpha in a linear fashion over those [inaudible]. So, the idea is that by looking at the first, the second and the third and keeping on going, you've exhausted some of that control you have over the false positive. I told you it wasn't going to be quick. And it's probably not coherent. But, you know, that's what alpha spending is [inaudible] about.

>> I think we got the idea. No, thank you very much. That was great. Leticia, you're running the Renovate trial, which again looks at noninvasive ventilation versus high-flow oxygenation. Do you have a feel for what actually caused the difference in this trial? What is the mechanism by which CPAP seemed to be more effective than high-flow?

>> Honestly, it's very -- I am trying here to reconcile this new data with the previous literature. Like, for example, in the Feralli trial, we saw that the high-flows seemed to be better for hypoxemic acute respiratory failure. Now that we have these results, that now we are talking about COVID, and maybe there is some specificity to it. What I can say is that Renovate is running. The recruitment is above the expected. And Renovate was not designed to be a COVID trial. But it kind of started, and then the pandemic came. So, it completely became a COVID trial. So, we're soon going to have our results. Renovate is comparing two levels of pressure, so BiPAP versus high-flow and enforced subset of acute respiratory failure. But, as I said, COVID completely dominated it. So, it's going to be interesting to look at our data. But, in a way, our data is complimentary to Recovery-RS, because we're not testing against low-flow. One comment I would like to make is still on the crossover in the low-flow. In the very nice comment that Manu made that it didn't -- it does not influence the inferential analysis that we make when we compare CPAP versus low-flow, because the crossover towards -- makes the effect go towards the null, and we still see an effect. But the crossover does influence what to -- the inferential analysis we make on the high-flow versus low-flow. Because, well, we didn't find any difference. Could it be because of the -- could it be because of the crossover? Could it be just bad luck? Because we know that trials -- that there is some uncertainty around the point estimate that we get in a trial. So, I think that the information Recovery Respiratory Support brings to the community is super important. But there are still some aspects that we need to refine.

>> Rob, you're muted at the moment.

>> Thank you very much. Janice, fantastic results tonight. But in low- and middle-income countries, the Tocilizumab result and now the result tonight demonstrating that the CPAP seems to be more effective. Does this create issues for less well-resourced areas, where there now may be a struggle to try and get these interventions, and get them to the populations that need them?

>> Thank you. It's a very good question. I think now that we're, you know, advancing with the -- with the care package, right, so if you think about the oxygen corticosteroids, IL-6 antagonists and now CPAP, right, that this is something that making access and availability is really important for WHO to do for low-/middle-income countries and member states. So, that's kind of like if we believe these are effective interventions, then we want to make it -- the access available. I think where low-/middle-income countries are -- may strug -- are, you know, are challenged in those things that we're trying to overcome. One is, like I said before, the access to oxygen. So, really access to high pressure oxygen, sufficient levels of oxygen, in order to run these sort of interventions. The second is in skills and workforce, right, that the training and the protocols -- simple protocols that are easy to use, SOPs around titration, starting interface management, all those details that you left up to the clinician to do in the hospitals in this study, that we would have to actually provide those kinds of tools for those clinicians and other technicians that are setting up devices, that are maintaining devices, that are, you know, decontaminating devices, all that package of the end-to-end kind of comprehensive care needs to be provided. And that's what's -- that's -- that is what we would aim to do for patients and clinicians in low-/middle-income countries. So, it's doable. It's a challenge. But I think we have to aim to get lifesaving interventions to the right places, to the places with all the tools around them in order to make it successful.

>> So, we've had a tweet from Anita Simons saying that these results make a good argument for having respiratory HDUs. Steven, you've touched on this earlier. Should we now start to institute respiratory HDUs? And, if so, who should be staffing them?

>> It's a great question, Rob. This is something that the Intensive Care Society and the British Thoracic Society have given some thought to during the pandemic. We do indeed wholeheartedly support the development of respiratory HDUs, respiratory enhanced care units. Respiratory support units is probably the best term, because it includes really all the therapies that we're talking about here. But I think the key thing is exactly what Janice just said. Whether it's low-/middle-income countries or indeed develop -- more developed countries, such as in the UK and Europe, we need the proper staffing, the proper protocols, the proper equipment, the proper safety features to make sure that these units run well and run efficiently. But there's no reason why they cannot run well, they cannot reduce the need for intensive care admissions, and they cannot get -- bring great outcomes to patients. So, I think it is something that in the UK and Europe we should think about seriously, and we should be rapidly expanding. One caveat, though, that I just wanted to sort of bring in, which is probably my own, again, interpretation of the results. It's probably more about what the results do not tell us. In my understanding, I think the results do not tell us that leaving someone on CPAP for 14 days, if they're not improving, is safe to do so. I think that if -- you know this -- the trial and its interpretation still requires good clinical acumen. Which means that CPAP may be safe to use, may be more beneficial than high-flow nasal oxygen, more beneficial than controlled oxygen therapy, but is not safe to be used without appropriate clinical knowledge and skill. In that if a patient is not improving, deteriorating on that therapy, the patient should still be escalated to the next therapy, as demonstrated in the real world by the relatively short time to intubation in the CPAP group of 2.2 days, which is -- which surprised me as well. But it's a sign that the clinicians conducting the trial were watching their patients very well I think.

>> John, a question for you. It's a question that arises frequently, and it's one I certainly do not know the answer to. A couple of years ago, we wouldn't have had a scenario where we would have had a preprint, a press release, a livestream all in 24 hours, and people start to implement the results of that trial. As a methodologist, has science changed significantly in the last 18 months that we should now be doing this? Or should we be waiting for the peer reviewed article to come out, and the pressures that that produce -- leaves on everybody else not instituting a therapy which appears to be effective?

>> I -- I'm not -- that that is a very difficult question to answer. You know, in the length of time it takes to get scientific results out is disgracefully long. And we haven't really changed that, you know, in the run up to COVID, in the decade before COVID hit us. I mean, that's a complex issue. You know, scientific publication is an extremely lucrative business. And, you know, people make serious amounts of money out of it. And that's probably why. It is very important to get results out quickly. But it's massively important that they're accurate. And, I mean, I've been doing trials for a long time. And I run a trials unit. And you make mistakes when you rush to do things. There's no question about that. I've been involved with a number of trials where we've been forced to publish -- you know, effectively publicize the results at a conference. And then we've had to retract safety findings and efficacy findings, because when the full dataset [inaudible] it's actually been triple checked, you discover that there's errors. I think the worst aspect of it is people use the preprints and, you know, the rapid publication to deliberately, you know, put dodgy results into the public domain. And we have to be very vigilant about that. I mean, some people are not averse to using preprints as a way of generating a kind of publicity stunt to make it more likely that the journal is then going to publish their finding. And I think that needs to be very vigorously [inaudible]. But, yeah, we need to do better. But we've got to be careful we don't -- we don't sacrifice quality and accuracy by doing that.

>> [inaudible] Danny and Gavin, [inaudible] bring you in on this as well, because, clearly, once you have this information to hand, you're as keen as can be to get this information into the hands of clinicians to improve patient care, yet it does create this difficulty. Maybe you could comment on that?

>> Yes. I mean, I think, you know, John puts forward some, you know, very wise, you know, notes of caution. I guess one of the things that we've been able to do, by being a relatively large clinical trial unit, is that we've been able to divert substantial resource, both in the initial setup of the trial and, you know, through this analysis phase. And, you know, if they're listening, our, you know, statisticians, Sweehan He [phonetic], Chen Ji, Ranjit Lall and Nigel Stallard, who have been overseeing this. We've had five statisticians working pretty flat out on this. And we believe that we've, you know, delivered to the audience tonight and in the preprint, you know, accurate and checked data. But I think, you know, nevertheless, you know, the richness of a peer review process, you know, is, you know, bringing in those extra aspects of external scrutiny. But I think, you know, when you hear a result and you see, you know, places running out of oxygen, the burden of disease, you know, globally, you know, it becomes a question of how long do you sit on, you know, that result before you share it, you know, with [inaudible] and all, with the wider community? And I guess, you know, we took the decision as the Recovery and other trials have done to put it out there. And hopefully, through forward [inaudible], you know, that I think, you know, bring in the aspects of different opinions and perspectives that, you know, hopefully, enables people tuned in to, you know, consider it, you know, in a balanced perspective.

>> So, Danny, the benefits of early dissemination outweigh the risks of perhaps getting something wrong in what looks to be a very robust trial, and those risks are probably quite low?

>> Yeah, I think that's right. I mean, my -- I really struggle with this. And my default position is that the peer review process has ended up the way it has done, even though there's room for improvement, because it's actually quite a robust process for, you know, this type of research. But, at the same time, whenever you see a result that, you know, might imply that there's a number needed to treat of 12, it's hard to sit on that. And I think the risk is vastly mitigated, as Gavin alludes to, whenever you can see an accredited expert trials unit involved. You know? And I think that adds a huge amount of reassurance to the value of the results. But, you know, I think this question is by no means answered. And I think -- I guess the point that Gavin has already alluded to this is that these are preliminary. And, you know, we think these will impact care, potentially. But we, also, need to highlight that these have not been subject to peer review. And, you know, we just want to assign a word of caution. And as Gavin said, you know, we just need to review them in balance, even though we are confident that the data is accurate.

>> Okay, thank you. We're running way over, just because there's too much to get through. A quick question. Manu, on this theme, you've been involved with the WHO, with the IL-6 antagonist meta-analysis. How long will it take, typically, for this sort of information to feed through now into the meta-analysis, and then through to the guideline group?

>> I think I can start answering, but Janice is the best person for this. I'll say that the protocol for the IL-6 meta-analysis, what we are talking about is a prospective meta-analysis. It involves getting all the trialists around a table, and developing a protocol with them, identifying the outcomes of interest, getting a harmonized dataset, analyzing and then publishing. So, it's a huge amount of work. And it involves mostly WHO doing all the work, and some of us, you know, getting to coordinate the kind of activity that goes around it. So, it is probably the best effort I'd probably say eight to 10 weeks of full on before you get close to a product that is viable to circulate. And this discussion about preprint reminded us of our discussions on the IL-6 prospective meta-analysis, where we were -- we were in two minds as to what exactly to do. But JAMA was very kind of helpful in trying to steer us as to how quickly they can turn this evidence around. Probably Janice might want to add something.

>> Thanks, Manu. Just a few things in there. I think from WHO perspective, one of the essential things for writing -- us publishing a guideline was that the data in forming the guideline was in the public space. And, traditionally, that meant in a peer reviewed journal. But the pandemic, like every -- like we know, kind of changed all that. Right? Because now there was like preprint availability. And from -- and from -- investigators are doing quality studies, then we were being like the colleagues here saying -- you know, like reaching out to us, in which we were very grateful for, saying, "We will have these study results at this time." That allowed us to begin to prepare our guideline development process simultaneously. And we had already been looking doing a systematic review on noninvasive and high-flow, and these sorts of interventions, for COVID. And, so, when we heard from the colleagues here, the investigators, that you soon will have full results, then we knew, okay, hold on, we'll get these results, we'll finish the systematic review, update it so that our estimates and everything that we show is the most up to date, but we had everything already in line. So, for this type of recommendation, as opposed to the -- the IL-6, which we were kind of working with investigators that still had ongoing trials that were just finishing and in progress -- in process, for this one, we hope we could finish within six weeks. We would go through our independent methodology team that will look at the data, grade the data, do the whole process, you know, keep in close contact to ensure that the data isn't changing, right, that as you guys proceed with your manuscript processing. And then -- and then get it out into the public space. So, we try to keep up as best as we can with the -- with the context that we -- because we all want the same thing, right, to get out a good, safe, transparent, quality recommendation for public use.

>> Thank you. So, we're going to finish with a couple of questions. Keith, haven't heard much from you yet. I haven't directed much to you at the moment. We had a comment about generalizability. So, the trial included those who were suitable for escalation to intubation. Will the results be generalizable to those who aren't suitable for escalation to intubation? And, in particular, I'm thinking, unfortunately, we're seeing a number of pregnant women coming through now who are getting quite sick. They weren't in this trial.

>> I think the reason for excluding pregnant women was about insurance, as is typically in clinical trials. You know, there will always be a degree of clinical judgment as to the generalizability of the results to both your setting and the individual patient that you're seeing in front of you. I guess the group that we're most cognizant of, and concerned that people might seek to generalize it, is those where, you know, that a decision is made that they're not suitable for [inaudible]. I don't -- I don't think our results help in that group of patients. And I think we do need further research for that particular group to understand how best to provide treatment to those.

>> Great, thank you. And, Bronwyn, one comment here reading from the chat function from Thomas Zacmadi [phonetic]. Do you have any processed data on CPAP delivery? Did you -- did you record any of that, do you know?

>> Yeah. And that would have been -- it's a great question, because it helps us to, I suppose, understand how the intervention was delivered, and whether clinician factors in terms of skill and experience and delivery, whether that actually impacts on improved outcomes. We don't have a huge volume of data around that, for obviously practical reasons. It would be lovely to collect that kind of information in this type of -- of this type of trial with these interventions, because we know that the skill and experience of clinicians delivering CPAP noninvasive ventilation in other contexts is hugely informative around how patients tolerate the intervention, how they receive it, and perhaps then how they -- how they do in outcome. So, unfortunately, for practical reasons, we don't in this instance. But I think it is important -- when we're talking about things like respiratory support units, and one of the benefits there is that you then end up with a more concentrated cohort of trained and experienced clinical staff, who can then specialize in delivering these interventions. So, I think moving forward that would be one of the benefits around perhaps concentrating delivery of these interventions into particular clinical areas. And, certainly, if we had the opportunity, we would collect that kind of data. But the circumstances didn't allow that at the time.

>> Thank you, Bronwyn. We are so far over time. Normally, I would finish by going around our panelists and asking for last thoughts. I don't have the opportunity to do that at this point. What I will do is quickly ask either Danny or Gavin just for a final comment. Anything you want to say. Perhaps what the -- what clinicians should do tomorrow when they go into work.

>> You know, I've -- Gavin, do you want to kick off first, and then I'll --

>> You can go first, and then I'll have another comment.

>> Well, no. So, I think I would say that CPAP is probably the treatment of choice if you're considering increasing respiratory support. I would encourage people not to use empirical therapies and randomized to whatever trial is available for their patients, rather than believe [inaudible]. And, again, you know, not directly to your question, but just to saying, you know, so much thanks to absolutely everyone involved at all levels in delivering the trial.

>> And I think, you know, Danny's covered what I was hoping would be my, you know, closing comments, thank our co-investigators, colleagues who work at TTU, patients and families most importantly, you know, who contributed to the trial, and the National Institute for Health Research, who funded the trial and provided the infrastructure to help us deliver it.