Dr. Derek Angus

Disclaimer: At UPMC HealthBeat, we strive to provide the most up-to-date facts in our stories when we publish them. We also make updates to our content as information changes. However, education about COVID-19 can shift quickly based on new data, emerging variants, or other factors. The information in this story was accurate as of its publish date. We also encourage you to visit other reliable websites for updated information, including the Centers for Disease Control and Prevention (CDC), Food and Drug Administration (FDA), and your state and local governments. 

Dr. Derek Angus, Professor and Chair, Critical Care Medicine, UPMC discusses Remap COVID-19. It’s a novel clinical trial underway to determine which therapies or combination of therapies work best in treating patients diagnosed with the virus.

Never Miss a Beat!

Get Healthy Tips Sent to Your Phone!

Message and data rates may apply. Text the word STOP to opt out and HELP for help. Click here to view the privacy and terms.

Read The Full Podcast Transcript

Tonia: This podcast is for informational and educational purposes only. It is not medical care or advice. Clinicians should rely on their own medical judgments when advising their patients. Patients in need of medical care should consult their personal care provider.

Tonia: It’s a novel approach, working to fast-track testing for COVID-19 therapies. Hi, I’m Tonia Caruso. Welcome to the UPMC HealthBeat podcast. Our guest today is Dr. Derek Angus. He is a professor and chair of Critical Care Medicine at UPMC and the University of Pittsburgh, and he is on a team leading the way for fast-tracking the testing of COVID-19 therapies. Dr. Angus, thank you so much for joining us.

Dr. Angus: Oh, you’re most welcome, Tonia.

Tonia: So, this is really a revolutionary approach that began some time ago. Explain to people what the REMAP COVID-19 program is.

Dr. Angus: Sure. So, I had for a long time been interested in trying to think about smarter ways of doing clinical trials, and indeed had been part of a group that had tried to put together at the international level a way of studying patients that get sick in pandemics using a design called REMAP. REMAP is, it’s part of a family of a new kind of trial called adaptive platform trials. So, normally in trials, things you’ve all heard about, there are, for example, just two arms. Everyone gets randomized to a sugar pill or the active treatment, the so-called placebo or treatment. And then the trial just runs and then you wait till the end to find out which is best. And people don’t like that for lots of reasons. They’re not interested in being a guinea pig. It feels like it’s research, not treatment. And so, we pretty much started by saying, all the sacred cows of clinical research come out of the room, let’s start from scratch and think through what’s the problem we’re trying to solve. And we wanted a design that did not make clinicians or patients feel uncomfortable about participating. Especially something like COVID-19, where it’s a pandemic and people are absolutely feeling a huge imperative to try to do the best thing for the patient. Yet at the same time, we knew there would be no active treatment. And so, we have lots of potential candidates. What would be the most efficient way to test the candidate drugs while also treating the patient as well as possible? So, the REMAP design is a so-called adaptive trial that actually tests entire recipes of treatments, and only a very small proportion of patients would just be in the so-called usual care or standard care option. And then, the word adaptive is specifically because the trial actually learns as it goes. So, if you can imagine you were going to a restaurant, and it gave you three choices for the starter, three choices for the main course, and three choices for dessert. And you had to order everything at one go. There’s many combinations there; that’s actually three times three times three is 27 different recipes. What’s your favorite combination? We don’t know. In this design, we’re essentially testing 27 different combinations, and only one in 27 is usual care. So it’s not 50/50. Furthermore, as soon as these other interventions, these other recipes, are doing better, we actually change the roll of the dice so that patients automatically start to be offered the better and better-working treatments.

Tonia: So, how many different kinds of treatments are there? Are they all drugs? Are they other forms of therapy? Explain to people what some of these treatments are.

Dr. Angus: So, the whole trial is already testing many different combinations. In fact, across the world, this trial that we’re running already has, this might be hard to believe, 1,600 different combinations. That’s because when you’re testing several things simultaneously, the total number of combinations rapidly gets huge. What we started with in Pittsburgh, in part because of just drug availability, was of course the hydroxychloroquine that you’ve heard about, along with different combinations of steroids. That already becomes about eight recipes. We’re rolling in, as you’ve heard, convalescent plasma. As soon as patients are getting convalescent plasma or not, plus all these other combinations, that immediately doubles the number of options up to 16 options. And then we have several other targeted immune modulators. We’re also launching different ways of actually looking at the coagulation in the blood because that’s increasingly looking to be very important in the disease. And so we expect that we will be up to close to about 100 different treatment options at UPMC within about the next week to 10 days or so.

Tonia: Because these numbers are so vast, how do you know what’s what? Give folks a sense about how you come to conclusions about what’s working and what’s not.

Dr. Angus: Right. So, at the heart of it all is still very ordinary-looking statistics, although they are so-called Bayesian statistics, which is maybe too much detail to get into. But each individual recipe and each individual drug is still being tested statistically within the trial. Everything gets simulated ahead of time in computer software that models the potential outcomes. And so, then the trial keeps running until we hit a threshold. So, we don’t stop testing and intervention until we’re 99 percent sure that it works. And I mean that literally, 99 — it’s actually a 99 percent automatic statistical trigger that gets simply hit by when the software says, “Aha! You’ve now gone through through this threshold.” That piece of the trial stops. And there’s an announcement made that says, “Aha, we’ve now worked out that this particular drug should be given to everyone,” and we go, OK. So that’s an announcement. Meanwhile, all the other combinations still keep on getting tested.

Tonia: How key is it, and how important, that AI and machine learning are a part of this?

Dr. Angus: Right. So, the inner guts of the statistics of this thing is actually something called a multi-armed bandit problem, which is an old machine-learning problem. It’s actually the most basic model in so-called reinforcement learning, which is a branch of machine learning that tries to learn as you go. That’s at the heart of this entire statistical engine. That, by the way, is also at the heart of what you all encounter every day when you’re on Google or Amazon. When you go on Amazon and you notice, how come I ordered one thing and now I’m getting ads shown to me that seem to be linked to that? They use very similar mathematical algorithms, and that’s essentially what we’re doing.

Tonia: And you’ve been on the forefront of this for a long time. It was really back in the time of H1N1 that you and other world leaders, health leaders, began to take a closer look at this.

Dr. Angus: Right. So, as a critical care physician, I’d long been interested in sepsis and in acute respiratory distress syndrome, which are basically the syndromes that define severe H1N1, define severe COVID-19 disease. In fact, most viral pneumonia pandemics present with the pointy end of the wedge, or the tip of the iceberg, are these patients who came into ICUs with acute respiratory failure, shock, and so on. So, we had taken care of many H1N1 patients, and we had noticed at the time that when we were taking care of these patients, we had no active treatments for H1N1. And then we said, well, we should do trials. And so, we and many colleagues around the world scrambled to try to get what I would call traditional trials up and running. And as soon as we got the trial up and running, H1N1 had left our region and had sort of rolled to another part of the world. So, then, you’re right, I and a bunch of people, it was not just me, we got together after H1N1. And first of all, it was a big international community that went to Toronto, and we sort of rubbed our chins and we said, “Oh, what can we do better?” And we started then at that time, we had invited some statisticians who had been working in the cancer space and had really been advancing this idea of these Bayesian adaptive trials to come to the meeting. And we knew each other a little bit, but not too well. We all got in a room, and we just started kicking ideas around. And that was the beginning of the nexus of an idea for a sort of an adaptive trial that could be waiting in place for pandemics. And that became enough of a nexus of an idea that we then got some funding from the British embassy, actually a program that stimulates collaboration between British and American scientists, to host a follow-up meeting in Pittsburgh. In that meeting, again, a bunch of people came in from around the world, and we really hammered on the problem for a couple of days. And that was probably the birth of the REMAP idea. It takes a long time. At that meeting was Jeremy Ferrar, who then went on to run the Welcome Trust, and he was instrumental in convincing the European Union to put out a so-called RFA, a request for applications, for a funding announcement. And so, we, as part of a consortium, applied to that, and we got a large grant to essentially be paid to try to, oh, wow, try to turn this, what was just an idea, into an actual design.

Find Out How UPMC Primary Care Is Keeping You Safe

Tonia: How important is it that so many countries around the world are involved in this?

Dr. Angus: So, COVID-19 is an international traveler. We’ve all watched that virus travel around the world. It became very clear even in prior pandemics, the speed with which these viruses travel around the world. And so, we’ve always felt it was an important philosophy that this trial run is an international trial. If we successfully contain our surge in the United States, we’re not going to have any cases here. You know, the total number of cases will die down, and then we’ll just have like small outbreaks or brushfires. And that’s what everyone is trying to do. What that means is we want to get rid of this disease. We’re glad that the total number of cases might drop, but we still want to pounce on every case possible. So, you almost want a spiderweb across the globe of sites that are ready to enroll any patient possible so that you can learn from as many patients as possible. So, we think an international approach is essential. Having said that, an international approach is complicated. The virus crosses borders, but funding agencies, investigators, institutions, legal contracts, like almost nothing else, crosses borders very easily. And so, it’s definitely been a challenge that the community is incredibly willing, the investigators are incredibly well-motivated, but there’s no question that pulling off something at an international level involves a huge amount of overcoming logistical burdens.

Tonia: How soon did you begin thinking of this and rolling plans into place once UPMC began to prepare for the pandemic?

Dr. Angus: Right. So, there were two things going on here. First of all, we had really helped be instrumental in leading the design of this international REMAP design specifically for pandemics. Secondly, we at Pittsburgh had already started to invest in the idea of REMAP designs not just for pandemics, but actually as part of a learning health system. And so, we had already begun to chip away at this idea of taking the entire trial architecture and putting it inside the electronic health record. That, to us, seemed to be the ultimate solution for running these trials. Then they run at the entire community level. So, in basically around the beginning of March, I think it was, when UPMC realized that along with all of the United States, hey, cases are going to hit the US, let’s start getting ready. On the day that we were forming planning committees to get ready to provide clinical care, we immediately started thinking we should bring this REMAP program here for COVID-19 and actually start working with our colleagues in IT, et cetera, to nest the entire trial design inside our electronic health record. And then, who would do the consenting? Who would be the clinical researchers? There was a worry about not having enough PPE, we wanted to keep the PPE for the clinicians, so we had to build an entire system from scratch to put the trial inside the electronic health record and then hook people up by video link to essentially a 24/7 telemedicine command center that we built from scratch over in Forbes Tower. So, all of that happened in, that went from zero to being up and running and RB-approved in about three or four weeks.

Tonia: The electronic health records, you’ve mentioned that a few times, and that’s key because that allows you to quickly enroll patients?

Dr. Angus: Yeah, so there’s a couple of issues about this. So, most clinical research in the country takes place in big academic medical centers. So, UPMC of course has big academic medical centers like here in Oakland, like Presby, et cetera. When people present with clinical problems to some of our community hospitals, we don’t have the infrastructure in those hospitals to do clinical research very much. So, when they’re there, they then need to be referred, they need to come into Oakland, et cetera. OK, there’s no time for that with a disease where minutes count. So, part of our build-out for this was also part of our overall plan to try to make sure that any patient hitting any part of the UPMC system with COVID-19 would all get the same care. So, we built out in the electronic record, essentially a screening system that captures the patient instantly, immediately finds out what information we need to know even for our reporting to the CDC, et cetera. And then right there, even if this patient is presenting to, say, our colleagues at McKeesport, they can just click on a link, write in the electronic health record after they’ve spoken to the patient or the family and they said, “Hey, we are trying to offer access to experimental therapies. Would you like to hear more?” And then as soon as they click on the link, that then activates a live feed of the patients or the family members’ cell phones numbers into our staff. And then we start up FaceTime and a video link to explain the whole study. That never existed before. And now that we’ve done it for COVID-19, we’re sort of thinking this is really the model for how we would really like to bring any UPMC innovation across the entire system.

Tonia: So, let’s talk more about the study. If I’m a patient and I’m in the hospital with COVID-19 and I’m on one of these therapies, but it’s not working, do you cut me off of that and move me to something else? How does this work in the hospital?

Dr. Angus: Right. So, any individual patient, once they’re put on their recipe, will follow that recipe. We won’t re-randomize to different recipes within one patient. Although, the recipe is tied to the patient’s course. So, for example, the moment you hit the hospital, the algorithm already thinks, well, if you do get sicker, and for example, go into respiratory failure, then we have some treatments that we would maybe only give at that point. The randomization to potentially receive those therapies is made instantly, and then if you do deteriorate, then there’s an automatic prompt in the electronic record that then starts you on that therapy. So, you might march through different therapies over time during your course, depending on the shape and course of your disease. But we can’t re-randomize within the patient’s course because it’s too short. This kind of adaptive platform trial design for chronic diseases can actually be used to think about changing a patient’s course in the disease, but that’s a disease that takes months or years. That’s not possible for a disease that only takes days.

Tonia: The silver lining is also a challenge here in that there aren’t as many COVID-19 patients in the UPMC health system as originally anticipated. So, how does that impact the study?

Dr. Angus: Yes. So, I run critical care for the system. So, I was very interested in doing two things simultaneously. One, launching a trial that could offer the best treatments as fast as possible to everyone who was sick. And two, trying to have as few sick people as possible. So, these two goals, as you suggest, they actually compete against each other. So, we are trying to make sure that every patient who is positive for COVID-19, we certainly approach. And some patients don’t want to hear about any experimental protocols. But we’re trying to make sure that we approach everyone and give an opportunity to everyone who wants to participate to be in the trial. And then secondly, that’s why this is tied into an international design. So, our data go along with the data from other centers to one large coordinating center that then looks at all the data worldwide. And then we can re-update those probabilities as frequently as weekly. What that means is that even if we’re not seeing that many patients here — let’s say we see a patient next week at Passavant who shows up — that person in Passavant is essentially benefiting from the rest of the trial’s learning that took place based on patients enrolled in London, in Sydney, in Amsterdam, and in Toronto.

Tonia: So, that all said, how soon will we know what’s working and what’s not working? Talk about that.

Dr. Angus: So, there’s two ways of looking at this. In terms of the raw numbers across these thousands of different combinations, on average, we expect to get to pretty strong answers about individual therapies with about 1,000 patients or so. If something has a very strong effect, it will only take a few hundred. If something has a more moderate effect, it could be 2,000 or 3,000. So then the question is, how long before you have several hundred patients? And that really depends on the burden of this disease. At the current rate that we’re enrolling, then we definitely expect to have answers coming this summer. You’re probably going to ask when in the summer, and I would say it depends. If the caseload goes down worldwide, then it will take longer. If we continue to see a lot of cases, then we should have answers early summer. If it slows down, I’m thinking late summer.

Tonia: And so, right now, you can’t say this treatment is working and this treatment isn’t.

Dr. Angus: So, first of all, although there’s a central algorithm that knows, and it’s getting the so-called unblinded data, and there’s a statistician that sort of presses the button on the algorithm to run and that person knows, all the data then gets shared with a so-called data safety and monitoring board, who are seeing all the raw data. But all the investigators have no idea what’s happening.

Tonia: So, how do you think, in the future, once we get past this, this REMAP-ing will impact clinical trials for conditions or illnesses of all kinds?

Dr. Angus: As someone who both takes care of patients and does clinical research, I have been frustrated for some time that we create these, almost as two separate worlds, the world of clinical research and the world of clinical care. They’re even run by separate organizations. You have Medicare and health insurance companies paying for the care, while you have the NIH paying for the research. So you have different people, different organizations. It looks similar, but they’re actually doing two separate jobs. And the problem about two doing two separate jobs is that everyone is either doing or learning. We’re very interested in trying to do both simultaneously. It’s called learning while doing. And that actually involves setting out two goals and caring about the perfect trade-off, because sometimes they’re in opposition, and actually trying to really be efficient at doing both at the same time. REMAP is one of the tools that we think enables truly learning while doing it. In other words, we don’t want it to be what we think is a false choice. We don’t want to either treat the patient or study the drug. We actually want to advance a philosophy where we use designs like REMAP so that we can really feel like we are treating the patient and learning at the same time. So, of course, this is what we care about, so we’re no doubt somewhat biased about it. But I definitely think of REMAP as absolutely one of the key elements of our toolkit for trying to advance a philosophy of learning while doing.

Tonia: Dr. Derek Angus, thank you so much for spending time with us today. It is cutting-edge research, and we cannot wait to hear about the results. Thanks so much for your time.

Dr. Angus: Thanks, always, Tonia.

Tonia: And thank you for joining us. I’m Tonia Caruso; this is UPMC HealthBeat.

About Infectious Diseases

If you have a disease caused by bacteria, fungi, parasite, or virus, the UPMC Center for Care of Infectious Diseases can help. Our team of experts is specially trained in the prevention and treatment of infectious diseases, including of HIV-AIDs, postsurgical and transplant infections, illnesses caused by international travel, and more. We research infectious diseases and participate in clinical trials to learn more and develop better treatment and prevention methods. Visit our website to find an expert near you.