Thank you so much, and thank you for the opportunity to speak. And I'm presenting today on behalf of all of the RemapCAP investigators, of which there are many. So these are the disclosures. There are many here, which is my fellowship support and the many global funders of RemapCAP. So I'll just start with a brief description of RemapCAP. So it's a randomized, embedded, multifactorial, adaptive platform trial in community-acquired pneumonia. And as a platform trial, it can evaluate multiple treatments for CAP simultaneously. So we began recruiting patients in 2016, but the trial was always designed to be able to adapt in the event of a respiratory pandemic. So in early 2020, we began recruiting patients with COVID-19 and evaluating interventions for COVID-19. So in our pandemic strata in RemapCAP, we enrolled patients with suspected or proved COVID-19. And we enrolled them in two states, a critically ill state, which is patients who are receiving organ support at the time of randomization, or a non-critically ill state, which is patients who aren't receiving organ support. And the focus of today's talk is patients in the critically ill state. So a patient who comes in with COVID-19 can be randomized into multiple domains, where domains are treatment groups. And you can see some examples here of the domains which are either ongoing or have been completed in RemapCAP. A patient can be randomized to multiple domains at the same time, depending on their eligibility. So they can receive an allocation to one intervention in each domain to which they're eligible. On average, I think patients are randomized to 1.8 domains in RemapCAP, but we've had single patients who've received allocations in up to seven domains at the same time. In Remap, we use a Bayesian design. So we do frequent analyses, and we continue randomizing in a domain until a pre-specified statistical trigger is met. So to date, RemapCAP has reported results from six domains, our corticosteroid domain, our immune modulation domain, our immunoglobulin domain, our antiviral domain, anticoagulation, and antiplatelets. So these domains, you can see here, recruited patients between March 2020 and June 2021. And what you can see here is that there was overlap in the recruitment for the majority of the domains, meaning that based on eligibility, patients could be randomized into multiple domains at the same time. Now, for those publications that I showed you, the primary outcome for COVID-19 patients in RemapCAP is the composite of in-hospital mortality and in survivors the duration of organ support to day 21. So it's an ordinal scale where anyone who dies in hospital is given a score of negative one, and otherwise survivors are given a score of zero to 21 based on the number of days free of organ support with higher numbers indicating a better outcome. So this is what those initial manuscripts showed. To the left of the dotted line indicates harm with the intervention or favoring the no treatment or standard care approach, and to the right indicates benefit. So you can see that for that initial primary outcome, we found harm with the antivirals, and you can see the clear benefit there with TOSI and SARI, the two interleukin-6 receptor antagonists. You can see that the credible intervals for corticosteroids there both cross the dotted line. In RemapCAP, we stopped our corticosteroid domain early before reaching a statistical trigger based on external evidence when recovery results became available. So we had a limited sample size to be able to detect an effect in that domain. We didn't find an effect in any of the other domains, and you can see there with therapeutic anticoagulation that that's primarily lying on the side of harm compared to standard care thromboprophylaxis. So what this, today's presentation is about is about the longer term outcomes from all of those domains. And what we know is that most randomized controlled trials in critically ill patients, and in particular in COVID-19, only assess short term outcomes like organ failure or day 28 mortality. But we know from qualitative work that actually longer term survival free of major disability and with an acceptable quality of life is actually more important to patients and really important to measure. So in REMAP-CAP, we've always specified these day 180 outcomes, so they are pre-specified outcomes. And so the aim of this analysis is to determine the effect of all those reported interventions on longer term mortality, disability, and quality of life. So in terms of the methods, so to be in our pandemic strata in REMAP-CAP, you must have suspected or proven COVID, and to be in our critically ill state, you've got to be receiving respiratory or cardiovascular organ support. Now, there are numerous exclusion criteria, the main one being that you're excluded if you've been in ICU for more than 48 hours before randomization. And there are, as you can imagine, many domain specific exclusion criteria, particularly related to things like hypersensitivities to the randomized interventions. So I don't have time in 15 minutes to talk you through all of the interventions, but we will go through them. But within each intervention, there are two or more components that a patient can be randomized to, so they receive one allocation in each domain to which they're eligible. So in REMAP-CAP, all patients are followed up to day 90, but following patients up beyond day 90 to day 180 is optional. And we only follow patients up beyond day 90 if they're in the critically ill state, so the patients who've been receiving organ support. So whether sites participated in the longer term follow up was a decision that was either made regionally or made locally at the site level based on things like funding, site resources, and regulatory approval. So in the six domains that we're talking about today, 88% of participating sites collected the day 180 outcomes. So for this analysis of longer term outcomes, our primary outcome is day 180 mortality with secondary outcomes of day 90 mortality, health-related quality of life and disability at day 180. So we measure health-related quality of life using the EQ5D5L. So this is a questionnaire with five questions that cover five dimensions, mobility, personal care, usual activities, pain and discomfort, and anxiety and depression. And the combination of a patient's responses are used to generate a utility score. So this is a score that ranges from negative 0.593 to one, but it's anchored at zero for death and one for perfect health. So negative scores represent states worse than death and higher scores indicate a higher quality of life. We measured disability using the World Health Organization Disability Assessment Schedule, or the WHODAS. We used the 12-item WHODAS. It has 12 questions about a patient's function. Each question has five options ranging from none to extreme or cannot do, and you can see a sample here. So it's then given a score out of 48, and that score is used to place patients into five categories of disability, ranging from no disability to complete disability. So with our Bayesian design, there's quite a complex statistical analysis, which I don't have the time to go into today, but just a couple of key points. The analysis plan was developed before we unblinded the day 180 outcomes. The P2Y12 inhibitor and aspirin groups were reported as a pooled antiplatelet groups, and the TOSI and SARI groups were reported as a pooled interleukin-6 receptor antagonist group, and that's because they met the trial-defined, predefined statistical criteria for equivalence. Importantly, our models adjust for location, so site nested in country. They also adjust for age, sex, and quite importantly, they adjust for time period. And for the health-related quality of life and disability results, we did multiple imputation where results weren't available for patients, and then we did a sensitivity analysis without imputation. So in terms of the results, so this was published online in JAMA in December and in print in the first edition this year, and there's a lot more results in the paper, so I'd encourage you to read it. So we had nearly 5,000 patients randomized into one or more of those six domains, and after removal of the patients who withdrew consent, we had 4,791 patients in this analysis. Now, they're from 197 sites in 14 countries, and you can see the sites on the map here. So there were sites in North America, Europe, the Middle East, Asia, and Oceania. Now, as I said, follow-up beyond day 90 wasn't compulsory, and so nearly 10% of these patients were randomized at sites that didn't complete the day 180 follow-up. They only followed their patients up to day 90, and that includes all sites in Nepal, India, and the United States, who only followed patients up to day 90. So there's lots of numbers on this slide, but just to show you that there are very differing numbers randomized into each domain. As I mentioned, we continue to randomize until we reach a statistical trigger. So we have day 90 outcomes available for 98.5% of patients. Now, that drops to 85.7% for day 180, and as I said, follow-up to day 180 wasn't compulsory, so there were sites not completing that. And then the mortality rate at day 180 overall was 36.9%, ranging between 30 and 40%, depending on the domain. So in terms of the characteristics of the patients, they were fairly typical of COVID-19 patients in ICU. So a median age of 60, predominantly male, an Apache score of 13, and 75% of the patients were receiving either non-invasive ventilation or ventilation at the time of randomization. So if we look at the effect of those interventions on mortality, what you can see in this forest plot here, so to the credible intervals to the left of the dashed line indicate benefit with the intervention compared to its control, and credible intervals to the right indicate harm. So you can see that the interleukin-6 receptor antagonists, they have an estimated mortality reduction at day 180 of nearly 8%, and the probability of superiority compared to no immune modulator is more than 99.9%. The antiplatelets have an estimated reduction in mortality of 3.5%, with a 95% probability that they are superior to no antiplatelets. And then if we move down, we can see that with both hydroxychloroquine and the combination antiviral therapy, there's an estimated more than 10% increase in mortality at day 180. And the probability of harm in this case is more than 96% compared to no antivirals. So if we look briefly at a couple of the Kaplan-Meier curves, so you can see here with the immune modulators, the blue line of the interleukin-6 receptor antagonists is clearly above both the no immune modulation and anakinra lines on the Kaplan-Meier. If we look at antiplatelets, you can see that the aqua-ish line of antiplatelets is above the no antiplatelet line. But really interestingly here, you can see that the curves don't appear to separate until day 30, and that may be one of the reasons why research to date hasn't shown benefit of antiplatelets when they're only measuring short-term outcomes. One of the other things we wanted to look at was the association between our short-term primary outcome and the longer-term day 180 mortality. And you can see from this graph here that there was a strong association between the two outcomes. And this was really reassuring because it showed us that we could use a short-term outcome to be able to rapidly make changes and rapidly make recommendations during the pandemic, but that those short-term outcomes are actually associated with longer-term outcomes. If we look at quality of life, again, there's lots of numbers on this slide, but the key point is this was really hard to do during a pandemic. So we actually have quality of life and disability measures on less than 50% of our survivors for multiple reasons. Often they couldn't obtain appropriate consent from the patient before they were discharged, in which case they then didn't have regulatory approval to contact the patient. Often they had the incorrect contact details and couldn't get onto the patient. So as I mentioned earlier, we did do multiple imputation where we didn't have those values and did a sensitivity analysis without imputation. So in terms of quality of life, the median utility score overall was 0.74. Now that is very typical of a critically ill cohort at six months after ICU, but significantly lower than what the population norm would be in a similar age group where the median would be above 0.9. The bar chart shows the percentage of patients with problems so the dark blue line at the bottom is no problems, but what you can see is that at six months, more than 50% of patients are reporting problems with mobility, more than 50% are reporting problems with usual activities and more than 50% are still reporting pain and discomfort at six months after randomization. If we look at the effect of the interventions, so in survivors, so again, to the right of the dashed line here indicates a higher quality of life, to the left indicates a lower quality of life. And you can see there that the antiplatelet group, that's lying on the beneficial side in survivors, so showing that patients who received antiplatelets have a higher quality of life than those who received no antiplatelets with a probability of superiority of 97%. You can see that all those other credible intervals there are crossing that dashed line other than at the bottom for lopinavir, ritonavir, where if a patient who was treated with lopinavir, ritonavir, those patients had a significantly lower quality of life than patients treated with no antivirals. Now, the benefit of the EQ5D is the utility score is anchored at zero for death, which means that patients who die can be given a score of zero and that means you can look at the combination of both death and quality of life in survivors, which is what we did here and what you can see is there that again, to the right of the dashed line is benefit and you can see there that both the pooled antiplatelet group and the interleukin-6 receptor antagonist group clearly lie to the beneficial side with probabilities of superiority over their control groups of more than 99% and again, lopinavir, ritonavir is still showing harm when you combine both death and quality of life in survivors. If we look at disability, so the median WHO-DAS score in survivors was 6.8, remembering this is a score out of 48 and again, this is typical of a non-COVID ICU cohort at six months and nearly 38% of patients were reporting moderate, severe or complete disability six months after randomization. If we look at the effect of the interventions, so it's switched around here, so benefit is now on the left. You can see here that all of the credible intervals cross that dashed line, but reassuringly, you can see that the point estimates for all of the immune modulators, both corticosteroid arms and antiplatelets all fall on the beneficial side, but again, with credible intervals that cross that dashed line. So this analysis has many limitations. Obviously, I've mentioned that we didn't mandate collection of data until day 180. There were substantial amounts of missing data for our quality of life and disability outcomes and importantly, there were some differences between the patients who we do have data on and those we don't have data on and whilst that's not in the randomized comparison, that could introduce a bias into our analyses. We don't have baseline quality of life and disability scores, which is obviously very hard to do in ICU, but while this was randomized, we actually don't know for sure that the baseline scores were balanced and so the differences that we see at six months could possibly be due to baseline imbalance. And the patients were randomized in 2020 and early 2021 and so these results relate to the effect of the interventions prior to widespread availability of vaccination and on prior variants and so whether they're relevant to current variants and with widespread vaccination is unclear. But in conclusion, we found that the IL-6 receptor antagonists and antiplatelets both had a high probability of improved patient outcomes at six months. Importantly, we found that increases in longer term survival, so the interventions that were improving survival, that wasn't associated with reduction in quality of life or increased disability and there's always a theoretical concern that where an intervention reduces mortality, that those survivors have a poor quality of life and that's not what we found in this case. And we also found that in-hospital treatment effects were consistent for most therapies, meaning that we could use, it's appropriate to use a short term outcome in a pandemic and that is associated with longer term outcomes. I'd like to thank our funders and all of our partners and happy to take questions after all the presentations. Thank you.

Remap-Cap trial

community-acquired pneumonia

COVID-19

interleukin-6 receptor antagonists

antiplatelets