IN THE INTERIM - ICECAP: THE RESULTS

Welcome to Berry's In the Interim.. podcast, where we explore the cutting edge of innovative clinical trial design for the pharmaceutical and medical industries, and so much more. Let's dive in.

Welcome everybody back to In the Interim... I'm your host, Scott Berry, and we've got a first here on In the Interim... We are going to talk about trial results. This will be a bit of a public discussion of trial results. The trial is the ICECAP trial. We're recording this episode ahead of time, and when the paper comes out, when it's published in JAMA, we hope to release this and allow an opportunity for the PIs of the trial to discuss the results and what it means about the trial. So I thought this would be a really nice opportunity.

Will Meurer, who's one of the guests today, and I talked about the design of the ICECAP trial. So go back to episode 63 and consume the design — a good 45 minutes on the design. We may touch on it for those that don't have the opportunity to do that, just to make sure we cover the key ideas.

My guests today: I'm joined by Will Meurer, who is a professor of emergency medicine and neurology at the University of Michigan and a practicing emergency physician. He also does some work with Berry Consultants as a medical and statistical scientist, and he is the contact PI and the IDE sponsor for ICECAP. Dr. Robert Silbergleit, who's a professor of emergency medicine, also at the University of Michigan Medical School, a practicing emergency physician, and a PI for the ICECAP trial. And Dr. Romer Geocadin, who is a professor of neurology, neurosurgery, anesthesiology, and critical care medicine at Johns Hopkins University School of Medicine, also a PI for the ICECAP trial. And Dr. Sharon Yeatts, who is a professor of biostatistics in the Department of Public Health Sciences at the Medical University of South Carolina. She's a co-PI of the Data Coordinating Center for the SIREN Network and a co-PI for the Data Coordinating Center for this trial, the ICECAP trial.

So welcome all to In the Interim and our discussion of ICECAP.

Thank you.

I will start by saying that I was involved in ICECAP. I was involved in helping design the trial. I'm a co-author for the paper being discussed today as well, but I will try to get good feedback from the team.

So Will, we did a whole episode, 45 minutes, on the design. For those that don't have the opportunity to go back to do that, can you give us an overview of the design so that when we talk about the results, it all kind of makes sense?

Yeah, I'll try to do this fairly briefly, and I would encourage you to go over to the other podcast to listen to a little more detail.

So out-of-hospital cardiac arrest is the disease. These are adults who are suddenly dropping dead. Sometimes it's from a blockage in their coronary arteries. Sometimes it could be from a blood clot in their lungs. But for whatever reason, they are suddenly dead. Somebody comes around, brings them back to life, and they don't quite wake up. So these are comatose survivors of cardiac arrest. Arrests happen outside the hospital. They come to the hospital. Their heart has been restarted, but their brain isn't working.

The study's objectives were to determine if we should cool patients and how long we should cool patients. The preliminary data that we had to guide our practice came from a couple of trials from back in 2002: cooling patients down to about 33 degrees Celsius, or 90 Fahrenheit, for 12 to 24 hours improved the number of patients who actually woke up. Some of the criticisms of those trials were that maybe this was just fever prevention, and that fever was harming the people in the control group.

So ICECAP was designed through a joint venture between the National Institutes of Health in the US, the Office of the Director, and the FDA. We had a planning grant called ADAPT-IT, and we went back and forth with Dr. Berry, Dr. Geocadin, Dr. Silbergleit, people from FDA, people from patient advocacy groups, and a variety of stakeholders to create the design.

The design that we arrived at was a 20-arm trial, so separately in both people with shockable and non-shockable rhythms. And if you're into clinical medicine and emergency medicine and cardiology and neurocritical care, shockable and non-shockable rhythms are: if your heart stops, can you use a defibrillator to restart it? That's often a sudden injury that maybe is going to have less severity. So shockable versus non-shockable was our sort of one measure of severity. But we thought maybe you might need to cool people longer if they had more of an injury. So we set up these two parallel cohorts of shockable and non-shockable patients with cardiac arrest who have been resuscitated. They're in the emergency department, we want to know if we should cool them and how long.

So we set up this trial to range cooling durations from potentially 6 to 72 hours. The reason we did not have a zero-hour cooling was because of the guidelines that were in place due to those landmark 2002 trials. At the time ICECAP was designed, it was recommended that you cool everybody. So our minimum cooling group was 6 hours, our maximum potential cooling group was 72.

But the trial was designed with some guardrails, in that the trial started out randomizing people to only 12, 24, and 48 hours, because that was where the majority of prior experience from clinical trials was. There was a trial in kids that went out to 48, and those two 2002 trials were 12 and 24.

The study was designed using a Bayesian response-adaptive randomization algorithm, and the outcome was the 90-day modified Rankin Scale using a weighted scale where if you were perfect, mRS of zero, you got all the points, 10. If you were dead or severely disabled — four, five, or six — you got zero points, and then there were incremental points in between. And the reason we did it that way was we wanted an outcome measure that both rewarded a duration for survival and also could reward the duration if the quality of survival was better, so that amongst survivors, those who had more complete recoveries would be informative to the model.

So the model took the information from the people within the trial and created a randomization vector for the next set of patients who were enrolled in the trial — separate vectors for both the shockable and non-shockable rhythms. Again, going back to the point where we weren't sure: we thought maybe the more severe people needed to be cooled longer, so the trial could find that, or it could find they needed to be cooled shorter. The design was quite flexible. Scott showed us designs. We said, "Oh, that looks good," or, "Oh, that doesn't look so good." One of the things is it built in guardrails, so it didn't go to the really long durations unless there was sufficient evidence that the duration response curve was sloping upwards. And it could have a maximum sample size of about 1,800 patients, and could potentially stop every 50 patients or so after 200 if overwhelming evidence in favor of the six-hour duration emerged.

That's it.

I guess for me, brief is a six-and-a-half-minute treatise. Anyway — for Will, that's brief.

Only a couple of things. So there's no stopping for what we would typically call success. You're looking at 10 different durations of cooling. We would be allocating patients to those durations, so there's no stopping — there IS stopping for what we call futility, but futility's kind of a strange word here. It's that if there's more than a 50% probability that the six-hour duration is the best, which is the shortest, it would then recommend stopping. And there's a strong duration response model that dictates the shape of the curve within that.

I feel like that's a good summary of the design. Anybody need to add anything? And by the way, I should mention, we assume people have the JAMA article. They can see that as we're talking about it. We don't have slides, but we're assuming that people have access to it. Anything else to add?

So what happened? Rob — and in the backdrop of this, Sharon and I were unblinded. We were making the design go statistically in the backdrop. We'll talk more about that. Will, Robert, and Romer were not. So Robert, what happened in the trial? What was the overall — how did it end?

So the goal of the trial was to identify this duration response curve, and our objectives were to find the place on the curve that was best, if there was a place on the curve that was best, or to see if the dose response looked completely flat. Our proposition going in was that a completely flat dose response curve implies — or one can infer from that — the inability to show any efficacy. That if we showed that the curve was rising in any way, even without a zero duration control, that that would imply that there was efficacy. And if we had a completely flat curve, then we wouldn't be able to imply any efficacy.

And that's what happened. We saw a completely flat dose response curve, that Will is demonstrating on his shirt. And so our conclusion from that is that there's no duration that was better than six hours, and that we can't infer that there's efficacy at all from this therapy, which is what we had hoped for.

I can go on to say — or I can let —

Well, let's come back to that. So officially, the adaptive design hit the trigger for stopping in non-shockable, which ended up being approximately 75 to 80% of the prevalence — non-shockable. And then in the shockable, it did not hit a trigger, but the DSMB recommended we stop in that as well, and that decision was agreed to, and hence accrual stopped. All the patients were followed through 90 days, and now you're looking at the results in JAMA of that.

That's right. And I think it's probably important to point out, because you said that it stops when there's a greater than 50% chance that the six-hour arm is the best arm. By what we meant by the best arm — we define the best arm as the shortest duration that didn't look like it was any better than a longer duration. So it wasn't the duration that looked like it had better outcomes than the other arms. It was the shortest duration that didn't look worse.

So the curve of this is allowed to be flat, it's allowed to be increasing and then flat, and then it can go down. It can be down the whole way, it can be up. It can't go up, down, and then up again. And so the optimal duration is the lowest duration that reaches the maximum level. And it's restricted within this curve.

Quickly, one quick point?

I think for clinicians who aren't as familiar with Bayesian statistics and probability, 50% may not sound like that much of a number, but just bear in mind, you only have 100% to spend across all 10 potentially of these durations, or effectively in this study, eight durations. Six had more probability than all seven other durations put together. And that was why the study ended.

I know we have inclusion/exclusion criteria, but who largely did we enroll? So Romer, tell us, tables one and table two — what do the patients look like, that we're going to draw the conclusions that we're going to come to here?

So the interesting thing about what we really planned for ICECAP was, we wanted a representation of the cardiac arrest patients in the United States, vis-à-vis the trials that were done prior. A key characteristic of this really is to capture that cross-section that includes not only patients who have cardiac arrest of cardiac etiology, but patients of cardiac arrest across all etiology. That's number one. So this is the reason why we did shockable and non-shockable.

But within that, we were able to actually sub-stratify what were the etiologies under the non-shockables, which is interesting because it captured another population that we thought was representative of the United States population, and that's the respiratory arrest — or said differently, overdose, which is very common, actually up to 20% in this group. So that is the key thing that was not previously well described in the existing literature of out-of-hospital cardiac arrest.

The other thing that was interesting in this is that also to capture the United States population and how we deal with cardiac arrest outside is the percentage of bystander CPR. Most of the studies in Europe, they have a bystander CPR of about 80 to 90%. In the United States, it's about 40%, and this actually impacts on the injury as it comes in. We should be striving to have a higher level of bystander CPR in the United States. But in order for us to really formulate a treatment, we need to actually capture what the degree of etiology is, what the degree of injury is coming into the emergency departments. So those are the key things I think that stand out.

The other thing which was also interesting, when you look at the previous studies, is there was a lot of male predominance in their population. But for us, for ICECAP, we actually got that difference closer in terms of the sex difference. So in a way, we captured what we wanted to capture, but there were a lot of surprises there that I think we could discuss. So those are, I think, the big picture in terms of table one and table two.

And the male predominance in the other studies — just so those who maybe aren't as deep into the weeds in this understand, it's not necessarily bias on the part of investigators that led things to be about 80% male. But inclusion criteria for some of the European studies were that it had to be a presumed cardiac etiology. And in terms of how that was applied, there could possibly be some bias in that. But whether it's individual-level bias or systemic-level bias, that was what they would find in those studies. They would have about 80% male sex.

And ICECAP was 58% in non-shockable and 66% in shockable.

Yes.

So the results have been alluded to a little bit here, Sharon, but as the statistician and sort of the Bayesian statistician — what are the results? How do you summarize the probabilistic results of what happened in the trial?

So as was mentioned before, we were focused on these two rhythm types, and so the analysis was done within a rhythm type, but the rhythms were allowed to borrow information from each other. So at the end of the day, we had just under 900 patients in the non-shockable rhythms and just under 300 patients in the shockable rhythms.

Among the non-shockable rhythms, the probability that the six-hour arm was the target was 0.51, effectively — so just over the threshold that we had specified for futility. In the smaller rhythm type, that probability was 0.49, so it hadn't quite met that threshold that we had specified, but was certainly consistent with what we saw in the non-shockable rhythm.

So one of the goals of this trial within each rhythm type was to identify the optimal duration, and the highest probability one became the six-hour arm in both. We also did have this measure of, quote-unquote, "efficacy" — that if the probability there was an increasing region of dose response, that would be analogous to benefit of cooling. That longer durations — it's demonstrated cooling is a good thing to do if 12 hours is longer than six. If 18 is better than 12, that's demonstration. And we had set up before the trial started a 97.5% threshold: if there's a 97.5% probability that there's an increasing duration response, this is a sign of efficacy. And we didn't get close to that. Largely this is 50/50, because it's 50% it's six, and if it's above six, then there's positive duration response. So we didn't really come close to demonstrating efficacy at the same time.

So what does it mean? Everybody's looking at this, but I'd love to get everybody's view on what this means. Robert, interpret this for us.

So I think that this means that the effect of hypothermia is not so robustly protective that in this broad audience of all comers, it was able to demonstrate a signal. We'd hoped it was. We thought we were doing things differently than some other people had in other trials that would maximize the efficacy. We didn't talk about the fact that patients needed to be cooled very quickly. One of the complaints about previous trials was that people were cooled so slowly that that's why hypothermia didn't work in some previous trials. And we used a very rapid cooling as an eligibility criterion, but nevertheless, in all comers it wasn't working.

I do tend to think that this trial is about how to cool and what ways of cooling might be protective, because the essence of cooling being a protective biological response is already established in various animal models and other human conditions. So I think that what we were hoping was that duration was going to be one of these very important parameters at proving this. I think that it didn't show that, and so at the end of the day, we're going to be very interested in a lot of our specialized subgroup analyses and looking for phenotypes of cardiac arrest, where what we do can capitalize on this known biological effect. But probably it's not everybody with the disease, at least at this point, based on what we found.

Was there anything — talking about subgroups, secondary endpoints — did you find anything interesting in any of that?

Yeah. So I would just like to add, because Robert mentioned the cooling: in designing this trial, we really wanted a stringent criterion to get the patient to a desired temperature level in the shortest possible time, so we elected four hours. But in doing that, we excluded those patients that we couldn't get to the target temperature at that period. In doing that, actually, we selected — or maybe inadvertently looked at — patients that were potentially, and this is a hypothesis moving forward right now, probably more injured. Because what that means from a neurological perspective is the brain has the ability to fight back, and shivering being that. So the cooling ability is easier if you lose that reflex in the brain, and that is actually an indirect representation of greater brain injury. But at the point we were designing this, we were just targeting faster cooling. But inadvertently, what happened, maybe as a hypothesis now, is that the subset of patients that we cooled fast were those patients that do not have this protective shivering that is really a manifestation of bigger brain injury. So that's a subset that we're interested in looking deeper into at this point.

To get back to your question about subgroups: the predefined subgroups that we looked at, at first cut, don't show any suggestions that there's a subgroup that benefited either. But we do have a couple that are looking — well, one big project that was looking at much more granular information, including advanced imaging, biosignals, and other parameters that can do a more elaborate phenotyping of disease populations, and that data is still being processed.

That's called PRECICECAP.

PRECICECAP, that's a study.

Will, what is your interpretation of the data?

In this large, relatively unselected population of cardiac arrest patients, duration of cooling was not associated with improved outcomes. However, I think one important thing to note is that our non-shockable group had a survival of about 20%. This is nearly double the survival in the hypothermia group of the Hyperion trial, that had fairly similar inclusion criteria. Sometimes clinicians like to think survival for non-shockable rhythms is zero. 20% is far north of zero. So I do think the diligent care our sites and our site investigators and the clinical staffs at the enrolling places provided is very important, and that we should not think that just because duration of cooling was not associated with improvements in the way that we did this trial, that there are not important things we can do for these individual patients and important things that we can learn in the future about cooling and other treatments for cardiac arrest.

I thought you were going to veer into the effect of hypothermia, therapeutic hypothermia, on that, but you didn't go there — that that 10% historical rate to the 20 was somehow six hours or more being part of the story of that. But then you went to the general care that these patients do have and stayed away from anything about hypothermia being related to that.

Yeah. I don't know the details of the Hyperion population well enough to speculate greatly, but I think our non-shockable patients were probably younger, right? Because we did have more overdose patients, and there will be subgroup analyses that look more in depth into our overdose/poisoning population. So we'll have important things to learn there. But I do think temperature control matters. Patient selection probably matters. We'll have to see in future studies. But yeah, in terms of depth of hypothermia, at least as we set up this trial with these inclusion criteria, these sites who were high-performing academic centers and community hospitals — this is what we saw. We did not see that hypothermia longer, out to 48 hours, could improve outcome.

And Romer, I didn't get your overall view on this.

So yes, let me just segue from what Will and Robert said. I think the interesting part here for me, taking care of these patients in the ICU, is that we know from the AHA historical data that the hospital discharge rate in this population is about 10.5%. So when we look at this trial, there's some improvement, although that was not really our target. Whether this bears out with what Will has said, because there's a lot of moving parts here — it's not just the cooling. It's the critical care, it's the response from the emergency department all the way to critical care, all the way to the decision-making that is wrapped into it. It's an entire package. So if you look at the AHA guidelines, it's a bundle that was applied on this patient. How did all of these things elevate the survival to about 20%? So there's something there that we need to tease out. Whether it's therapeutic hypothermia, as you mentioned, I don't know, because there's so many layers. Therapeutic hypothermia is just one of them.

I do want to touch on operations. We're going to come back to what this means. But this is a trial that did interims every 50 patients, ran Bayesian analyses, updated RAR, decision rules, interacted with a DSMB watching this. Operationally, Sharon, comments? How did this go? Any comments on the running of this trial?

Yeah — well, it was definitely different than anything that I had done before. I think that one of the key pieces of making sure that this worked the way that it was intended to work every 50 patients was making sure that someone from our statistical team was looking at the data. At the beginning, we thought almost weekly. We were looking to make sure that the mRSs were coming in, and we had complete data. We knew who was supposed to be included and at what time point, because we thought we were going to be doing these approximately monthly. It turned out to be every two months was more consistent. So as the course of the trial went on, we got more comfortable with that timing, that spacing spread out a little bit. It wasn't quite so time-consuming. But that was definitely something that we stayed on top of more than we would have in a traditional trial design.

I think another thing that maybe helped was that I didn't know what a reasonable expectation was for turning these around. So my assumption was that the turnaround time must be very quick, because we are randomizing patients while we are doing these analyses. And it turns out that because I didn't know that it was okay to take a couple of days to turn this around, we proved that we were able to do it within a business day. And that was a challenge because we didn't know when that 50th patient was going to come, right? It would've been much simpler if we had done it once a month or once every two months on some sort of calendar schedule.

And it's something I'm asked about with more complicated designs like this: was the model — was anything altered, or did this run as designed?

For the most part, this ran as designed, which I know maybe is not the norm. But in terms of the randomization of patients and the fitting of the model, we did not have hiccups in that sense that would require alterations. At least not that I can remember in hindsight.

Yeah, I mean, the same model that started it ran throughout.

It ran throughout, yeah.

Partly the very low accrual rate to the shockable, the results — it stopped at a point maybe slightly before the original stopping rule for shockable.

Correct.

But the algorithm ran. How about from the blinded PIs? Operationally, from your perspective, did the trial run pretty straightforwardly or not?

Well, sometimes people ask me to give career advice. I would say, with our first patient coming in on June 5th, 2020: if you're ever leading a trial that is at the intersection of emergency and critical care, if you can choose not to do it in the midst of a global pandemic, find a different time. It sure would've been nice to start the trial earlier.

Operationally, I think we did take care to try to avoid information regarding what was happening at the various durations. But unfortunately, due to some of the safety supervision things that we do, I knew that people were getting assigned to six hours, because sometimes sites had questions about it. So that did provide some information to me, and because I knew what the design was, I had to continue to keep an open mind. I had to be like, "Well, there can still be a pretty strong upslope to 12 to 18." And again, that was a little bit of a sort of personal existential challenge too.

I will say I feel a little bad in retrospect, because we needed to — because the trial was a little behind in recruitment, we needed to think about doing a renewal. And I asked you two to do a bunch of stuff like simulations and stuff that I know in retrospect now that you knew were pointless. And I have to salute both of you for your immensely good poker faces.

I think the big operational thing, though, is what was touched upon before. I think we felt that we could transform things in a way that we could get emergency medicine and critical care, for lack of a better term, to give — insert expletive here — about these patients, get devices on them early, take good, diligent care of them early. I think the concurrence of the pandemic and all the other things that were weighing on healthcare providers and families and patients — I don't feel that we gave as much of a push on implementation as perhaps we should have. And as Romer mentioned, we got a mixture of patients who were cooled really early by effective sites that were getting devices on patients quickly, and also by sites that just found people who were environmentally cold and whose brains were not fighting rewarming induced by the environment. Getting sites to work fast on these patients, I think, was our biggest operational challenge.

So you mentioned how this started enrolling in 2020. The results are now coming out here in 2026. The audience is looking at this paper. Will this trial change patient care? Robert, do you think patients will be treated differently because of this study?

So I think that this paper should reinforce equipoise. We didn't know before, and now this didn't take us into an area that we know a lot more. So it's not really going to change how I treat patients. I think that this says that temperature control for 24 hours, which is what we were doing — that we couldn't find something better, so I'm going to keep doing the same thing.

Do I worry that it's going to change people? I do think it is. One of the biggest problems we had, as Will just said, was that we were struggling against this prevailing attitude of clinical nihilism for these patients with poor prognoses in general. And there's a strong tendency for people to think that nothing we do matters. And I feel like this is going to be read as: that which I was doing before is no longer any good. And they're just going to stop doing anything for temperature control in these patients. And really, this trial didn't tell you anything about not doing anything for the patients. Everybody got temperature management in this trial. So you should keep doing what you were doing before. It doesn't convince people who weren't doing temperature management before that they ought to. But I am really concerned that people are going to just read the bottom line — neutral trial, temperature doesn't do anything — and just throw up their hands and do nothing. And I don't think it should be interpreted that way.

I do fear that too many people read clinical trials as a single-sentence conclusion at the bottom of the abstract, and that's the sum of what they think.

So I would just like to reinforce what Robert and Will said, because I think that as a neurointensivist and as an intensivist, we bear the brunt of this problem in the ICU. All of the complications that happen, and most importantly, the other thing here is culture — the attitude of the staff and the hospital towards these patients. And as Robert said, the nihilism is just so much. And one of the things that I think ICECAP was partially successful in, in some areas, was to actually change the culture. Because this trial is not only an ED trial, it's an ED, multiple departments, ICU, all the way to discharge collaboration.

The key point here: we may not have shown that longer durations were better. The key point here is cooling did not make these people worse. And the important thing is that there was no harm signal. I think the other important thing here that we need to take away is that despite there being no benefit with a longer duration, when you look at the actual survivorship vis-à-vis the previous trials, we were at par.

So now the concern, as Robert was saying, was that if we don't take care of these patients in the ICU and just give up, it becomes a self-fulfilling prophecy, which has been the overall recurring story in post-cardiac arrest — of the nihilism, and people will just give way to the old things: hypothermia doesn't work. But the last thing that I would like to say is that hypothermia in this sense was really sort of a surrogate of multiple layers of care and multifactorial collaboration that happened with this patient, that should not be taken away. Sure, longer durations may not be advisable, but the critical care that was already built in there should be continued. So we should not give up on these patients — that is my final message.

And I've got to — I don't know if I'm going to do this right, but can you see the sites on the back here?

Some of them.

Oh, no, I've got to get lower.

No, I don't think so. But there were 71 US sites.

I think you're going to say something about —

Well, I want to thank everybody at all the sites, all the patients, families, everybody who made this trial possible. Some of our lead enrolling sites: Pittsburgh, Adventist in Oregon, and Cooper in New Jersey — some of our sites that really changed their practice over time. I will highlight Yale, Dr. Rachel Beekman and the team there. They implemented an early cooling protocol that involved neurocritical care collaboration with emergency medicine early, and again, devices on patients early, getting temperature control initiated early. Again, unclear how long to do it, how deep to do it, based on the things that Dr. Silbergleit's mentioned. But they showed — and they've been working on some publications, they've had some other things out — they improved outcomes by taking care of these patients diligently early with early temperature control devices.

So I think in any clinical trial, there's heterogeneity. Again, we think of it at the patient level, but we have site-level heterogeneity as well. So it's possible, as Dr. Silbergleit was saying, that maybe this works in a different group of people. Maybe at sites that are taking an early approach. If we had had all sites like that, the whole trial, could the results have been different? Sure. Not the trial we did, not the data we have. I would say the sites that take care of patients and get temperature control on people early — that'll be an interesting thing to look at in some of our future secondary analyses.

So I look at the results of duration of cooling. Everybody in this trial was cooled. Everybody went through the procedures you talked about. Everybody's in to help these patients, and yet duration of cooling — there's no evidence that beyond six hours is beneficial. But largely, all three of you think that you did a good thing for the 1,100 patients you enrolled, that the general care that was part of this package was better with your expertise than how these patients do elsewhere.

What do you think this opens up? What are the new questions now? Is there another trial to run here? Is there a zero cooling? Is there lack of — above 37 prevention? What is the next trial to run, if there is a trial to run? Anybody want to touch that?

So I think that what Will probably wants to jump in here with is that there is another trial that is running in the pediatric cardiac arrest population, that is the pediatric ICECAP trial, that addresses some of the things we might have as regrets about how the adult trial ran. So there are some regrets. We created an adaptive strategy that so effectively closed out the longer durations that we never really got any patients on the longest durations. And while that was intentional — that was how it was designed to operate — maybe it was a bit extreme. Maybe we would've seen something different. I don't know. But the pediatric trial will allocate to all the durations, not just be driven entirely by the adaptive.

And potentially allocate to all.

Potentially. Yeah.

And it also does have a zero duration of additional cooling arm, which I don't think was one of the big limitations of the adult trial, but many people did. So there's another bite at the apple, with some issues addressed, that will give us some more direction. Whether it's positive or negative, it might give us some more direction as to what the next question is, as well as the phenotype analyses from PRECICECAP.

So I don't know what the next step is. I don't think that there's probably an appetite for funding a next trial right now. But there may be, once we get all the additional information from this trial and its follow-up trial and its studies, and that may generate the hypotheses we need. Because I do think that it is likely that there is some group that would benefit — but whether that's a practical group that we can identify and actually translate into practice in a meaningful way, there may not be.

So I would like to take off from that point, the last point that Robert said. I think one thing that we learned from this, doing this for a long time, is that it's good that we did this in ICECAP, that we actually did a nice cross-section and just showed the subsets. But I think we need to learn from stroke, we need to learn from breast cancer, we need to learn from how precision medicine is applied in the other clinical trials — which means that we need some kind of identifier in terms of the subset very early on in terms of phenotype. Because not all brain injuries are the same. Not all cardiac arrest comas are the same. And obviously we've learned now that we cannot save everybody across this condition. If we identify the specific subset that will potentially be still responsive to treatment, then it's almost like what stroke did for the longest time. They just did trials, and then they did thrombectomy, and thrombectomy represents a very small fraction of all the strokes, but they became very successful. So to me, that is the next level: do we need biomarkers very early on to identify a sweet spot that would give us a chance to save a fraction of these patients early on?

To Monday morning quarterback a little bit, Sharon — the design. You watched the design, you watched it go, in all of its detail. Do you like the design? Do you like what it did? Do you think it did a good job?

So I don't have the advantage of being part of all of the design discussions that happened a decade or more ago. I do think that the design did what we asked it to do. I to some extent think that it took a little longer than it might have needed to, to address the question, at least in the non-shockable group.

And I'm waiting to see how I feel about the outcome. The weighted modified Rankin Scale has been a little bit difficult, as we think through some of these secondary papers, in helping folks think through what that means and why we would do that as opposed to just a dichotomous endpoint that everybody understands and feels comfortable with. So I'm waiting to see how those discussions go. But I think the design did what we asked it to do. It ran the way that it was supposed to, and it did answer the question with many fewer patients than we had specified as our maximum. So from that standpoint, I'm happy.

I'd like to jump in there, though, with one of the things that we've just recently learned, which is that writing the primary manuscript for a trial like this that has a lot of moving pieces and is very complicated becomes a real challenge. Because the primary manuscript doesn't get longer, and people aren't going back to design papers. And so the number of things that need to be explained for reviewers that are going to look at it briefly, and presumably readers eventually — it's a real challenge. So better designs — I think it was a better design. It did what we wanted it to do, but trying to get it across to people succinctly in the primary paper was harder than we expected.

And we have an audience out there that's consuming the paper. We'd love to hear those things — whether the information is hard to understand, feels blah, all of those things. We'd love to hear that feedback.

Sorry, Will.

Oh yeah, no. I guess I would say, if we're going to do the Monday morning quarterbacking: there were lots of things that the two clinical PIs talked me off ledges for. One was, I really wanted to alter the four-hour rule, that people had to be cold within four hours, because it could've juiced our recruitment. Now, that midstream in a trial adds so many scientific questions, particularly with RAR in place. So every time I was like, "We're not enrolling fast enough. We need to do this," I'd stew about it for a month or something.

But the other thing that I was stewing about late was — to Sharon's point, I did feel that our futility rule maybe was a little too hard to trigger, that there could be definitely non-upsloping patterns that didn't quite achieve the exit criteria, so to speak. We saw that in the shockable rhythms. So that was one of my fears — that it's pretty obvious what the... Maybe this would've been obvious at 900 total patients, and that we didn't get a result out there sooner so that we could talk things through.

The upside of that is that these patients were getting good care from diligent teams, so the trial was not putting anybody in harm's way, with close monitoring of adverse events and everything like that. So these people were getting good care. The fact that it went on longer is sort of sad from a scientific perspective, so that we can't move on to the next things. But from an information perspective, the trial dataset's bigger, right? So there may be questions that we can answer in the secondary analyses that we wouldn't have been able to answer had the trial ended at, say, 800 patients. In the end, I don't really regret a thing. I think it did what it was supposed to do.

So let's end this then with a really hard question. In part I'm hearing this, that the trial could have stopped with this relatively same conclusion earlier — and I think it could have, you think it kind of answered the question — but on the other hand, it sounds like cooling is still in the cards for you. And so it's sort of an odd thing for me. So maybe the big question is: if this was you or a loved one, would you want to be cooled, and how long? If you were part of the inclusion/exclusion criteria, or a loved one. Robert?

I would still want to be cooled for 24 hours, which was our starting point. This didn't move me to longer durations, which is what I think the question was — should we be doing longer durations? And this does not tell me that we should.

Yes, I agree. I want to be cooled to 33, and I don't want the team to give up quickly, which I think is also another factor here. These patients deserve the full breadth of the critical care team, anywhere from five days to seven days at best, unless they're brain dead. So that's what I want to be done to me: full bore, five days, seven days, and get a real multidisciplinary team to take care of me.

Well?

You know, we did this trial for a reason, and I think the flat, or not up-sloping, line to me I feel is informative. So I would say I'd like a device on me early. I would like you to give me as much time to wake up as is logical at this stage of my life. Maybe when I'm older, I would want that number to be less. And yeah, set me to a normal temperature, but get the device on me early.

"Set you to a normal" — so is that therapeutic hypothermia?

Oh, no, no, no. Set me to like normal temperature. 37.

Okay. So what you want is something that we didn't even test in the trial.

Yep. Implied, right?

Okay, Sharon. Consuming the information, what do you do?

I'd like to cheat and answer this by saying that the only thing I know is that I want as much time as possible to wake up. And I would agree with Will. I would not want therapeutic hypothermia based on the results. But I've never treated a patient. Take that as you will.

All right. I appreciate y'all for doing this. It's a very interesting discussion. I can't wait to see what people have to say about the ICECAP trial. I think this is a conversation we may be talking about a bit more. And you all came on In the Interim, and we did many interims here, so it's a place you all know very well.

For the rest of you, until next time — thanks for joining us. We'll see you on the interim.