Please tune in to Episode 2 at Apple Podcast, Google Podcast, Amazon Music, Spotify, and Firstory

Transcripts:

Intro:

What are some differences between biostatisticians in hospitals and the pharma industry? What are some specific skill sets required for these jobs? And how can Bayesian framework be used in benefit-risk assessment of new medical products or interventions? Carl DiCasoli will answer all these questions for you.

Carl obtained his PhD in statistics from North Carolina State University and currently is working as the director of biostatistics at Sunovion Pharmaceuticals, where he acts as senior reviewer and lead biostatistician in clinical trials in neurology, designs innovative Bayesian adaptive phase 1 first-in-human single ascending dose study for a new compound at Sunovion and collaborates with Frontier Business Division within Sunovion to drive clinical trials for a digital therapeutics medical device to treat social anxiety.

Wait, what is that? I’ve also seen Carl’s name on the Piano World piano forums? And there are words on the street that instead of a statistician, he was almost a professional classical pianist? Let’s hear more about it in this episode.

Jocelyn: Welcome Carl, to our Biostatistics podcast. It’s really nice to have you with us here.

Carl: I’m also pleased to have you as well.

Jocelyn: Guys, I’m very excited to have Carl with us because he has a very interesting and not at all conventional background, but I’ll leave it to himself to talk about. So can you start by telling us a bit about your background and how you became interested in Biostatistics?

Carl: Yes. So after high school, I went to music school for professional piano. I was the music performance major at the New England Conservatory of Music. And I had a good time over there. And then after two years, I decided to pursue more of an academic trend. And I was trying to figure out what I wanted to do. I started first in computer science at the University of Central Florida. And then I found out about this very interesting field of biostatistics, which you are helping put drugs on the market, or research questions regarding health care. And to me, that was very enthralling and interesting. And hence, I decided to go in that direction.

Jocelyn: Oh, that’s interesting. So you were in the musical school, and then you switched to biostatistics. So how would you say that music or piano part affected your, I guess, your path in biostatistics?

Carl: Well, music is always something that I enjoy. It’s relaxing. It helps you look at a problem and understand it, the notes on the score and how to process what the composer is thinking. And in this case, it’s similar to statistics. You’re trying to process the problem and how you’re supposed to solve it. So in a way, it’s a very similar skill, but it’s different. I think the difference in music is it also involves one’s personal convictions. Although in the real world, you could think of it that way too. If you’re talking to a clinician, you’re talking about how to present something, it can also involve your personal way of doing it, your personal touch. The same way as you have an interpretation in music. And yeah, you’re starting to limit certain boundaries of what’s correct and what isn’t. But then there’s that leeway of what’s in the score and how to interpret it. The same thing with knowing the best way to present something to a clinician or to another function.

Jocelyn: I see. That’s a very interesting take. After we talk about your background, can you tell us a bit about what you’re doing currently?

Carl: Yeah. So right now, I’m at a company called Sunovion Pharmaceuticals. So they’re a Japanese company, part of the subsidiary of Sumitomo. So what it is is that the Sunovion is actually the neuroscience division, and they work strictly on neuroscience. And there’s SDP oncology, which used to be Boston Biomedical. They work strictly on oncology. And then you have, finally, the Sumitovant, which is sort of like this group of startup biotechs. And that’s another entity of it. So we concentrate on neuroscience. And I’m working on schizophrenia right now, the phase three trial with a new compound. And I’m also working on social anxiety with a new device, a medical device. And it’s virtual reality. So it’s supposed to help someone overcome their social anxieties by giving them challenges. And it’s an adaptive design according to how they respond to the challenges. And then you measure the LSAS, which is showing how they improve over time. And on top of that, I’m working on some early phase trials regarding schizophrenia, metabolic study in phase one, and also a PK study and a single ascending dose study in depression to try to find out if there’s a dose that’s effective yet, one that’s the maximum tolerated dose. So that’s the projects that I’m working on currently in Sunovion from a high level.

Jocelyn: I see. Can you tell us a bit more about the virtual reality part of your job? Because that sounds very new to me.

Carl: Yes. So what that is, is you’re looking at patients with social anxiety. There’s a device, kind of like the metaverse that we think the meta is doing. Patients wear this device and they’re given social challenges. And they figure out how they need to respond to these challenges. And they’re evaluated on that over a certain time period. And then you look at how they improve. And this is similar. This is like CBT therapy, cognitive behavioral therapy. And the whole point is to try to figure out, okay, is this working for them? It’s an intervention. And a company called BehaVR based in Nashville, Tennessee is the one who actually made the device and we’re testing it. At Sunovion, we’re testing it. And this is called the Frontier Business Office. So we’re hoping that it can work. We ran a feasibility study and we’re going to run a regular clinical trial with it. And hopefully this will be the new generation of therapies.

Jocelyn: That’s pretty cool. I’m wondering, what do you think is the difference between this kind of medical device as compared with the traditional medical device?

Carl: Well, this is more of a digital therapy. So it’s an adaptive digital therapy. Medical device could be anything that’s not a drug that affects the way the body works. So it could be anything like a women’s contraception, the way Bayer’s Essure was, or any of the cardiac stents and electronics they use. But this one is kind of more, almost like in virtual reality. There’s not many therapeutics. There’s some that are doing this, but not many on the market. This is kind of a new area. So I think this is how it’s different, the digital health aspect to it.

Jocelyn: That sounds really, really exciting. So you have worked in different places, and by different places, I mean cities and countries, and you also work with people from different countries. So I’m wondering, what would you say is the culture difference between all these countries and cities in the biostatistics industry?

Carl: Well, here’s the thing. The job is more or less the same. You’re interacting with cross-functional teams. You’re dealing with regulatory agencies like the FDA, the EMA, PMDA. I would say where the difference is, is perhaps, for example, working with Japan, there’s obviously cultural differences, for example. In Japan, they may not say yes or no as straightforwardly as they would in the US. So you have to adapt to things like that. I would say when I was in Bayer, I was working mostly with German colleagues and in Celgene with colleagues that were in Switzerland. But as you know, many people from all over Europe go to work in Switzerland. So it’s not necessarily Swiss people born in Switzerland. It’s people mostly from France, Italy, Germany, that sort of thing. So I would say there’s not much difference between that and the US people. But I would say regulatory-wise, there’s some differences in the agencies. But more or less, it’s the same. I think with Japan, there’s that cultural communication aspect which is a little bit different. And you have to get used to it. Because they may not tell you as straightforward no or yes, and you have to get used to it. But I think other than that, I believe more or less it’s the same. You operate mostly in a similar manner. That’s interesting. So in terms of the North American part of the industry, how do you think it will evolve in the coming years? Well, I think the North American part, the FDA has been more open to new methodology and getting drugs onto approval to market much quicker than in the past. So I think that we’ll have shorter times to approve, we’ll have more drugs on the market, and we’ll be able to have greater help for patients. And all these different methods, the Bayesian methods, the Estimand concept, which was not in existence several years ago, I just think that all these things will help us become more efficient to market with more novel drugs as time goes by.

Jocelyn: It’s very exciting to know that the industry is being more adaptive and acceptive and making the market more efficient. So are there any exciting projects you’re working on?

Carl: Yes, I would say these phase three projects are very interesting and exciting. These virtual reality devices, I would say the Bayesian approach to the SAT study is exciting because it’s sort of what was done in oncology, but it’s applied to neuroscience. It had been done previously by people at Roche and I had applied this to CNS and we could get a more accurate gauge of what the dose is. So in another old trial, I applied this method. The FDA, for example, told us to use 50. The SAG trial told us to use 100. The truth was that the best dose was around 75. But in the protocol, you didn’t actually test the 75 dose, but you were able to use modeling to logistic regression, like similar to MCRM in oncology, and be able to find that the most optimal dose was in between the lower dose the FDA told us to use many years ago and the actual dose that was proposed by the SAG study. So to me, that’s interesting.

Jocelyn: I see. You did mention a little bit about the Bayesian method. I noticed that you’re also, involved with the Bayesian Scientific Research Group. Can you tell us a bit more about that?

Carl: Yes, I’ve been involved with this Bayesian Scientific Research Group for a while. It’s involved in Bayesian benefit-risk. Benefit-risk is one of the very important key aspects when the FDA approves a drug. They’re trying to look at, does this drug offer more benefit than compared to the risk? For example, safety issues and the risk benefit, how effective it is, that sort of thing. There can always be some type of side effect, but can the drug benefit outweigh that? And when the FDA goes for approval, they’re always asking about that. There’s two types of benefit-risk. There’s quantitative and qualitative benefit-risk. In the quantitative benefit-risk, the frequentist methods may not be the best for this type of framework. The Bayesian methods can incorporate knowledge over time and you’re always updating priors and beliefs as they become adjusted. In addition, different audiences can have their opinions about what is more important or not. And if you have a Bayesian framework, you can use weighting and priors to do that. Hierarchical priors, for example, and weights with prior knowledge. If you have all these parameters in the frequentist setup and you’re estimating things like maximum likelihood estimation or method of moments, this becomes much more difficult and much more less tractable to find a solution, especially if you have all these parameters floating around. So just the Bayesian framework allows one to do this in a much more efficient and coordinated and interpretable manner to all parties involved.

Jocelyn: I see. In general, when we think about the Bayesian method, we think one of the benefits is that it can incorporate historical information. But I do notice that a lot of the time when I use the Bayesian method, I just use non-informative or minimally informative priors. So I’m wondering in practice, do you actually, I guess, elicit historical information or like informative priors a lot?

Carl: Yes. So if we have historical information, of course, it behooves you to use that. For example, one of the projects in my group was looking at the Voyager trial. The Voyager trial has thousands of patients over long periods of time. Of course, you want to update your learning based on that. If you have a trial, it’s a more rare disease type of thing and you don’t have much information, then you may not be able to do that, have informative priors. But there’s a lot of benefits having non-informative priors too. One may think, oh, the estimates are the same as the frequentist, but in reality, they’re not the same because the interpretation is different. It’s based on an actual probability rather than a confidence interval. You get an exact probability interval. You can be more confident that’s what it actually is. And just the Bayesian approach with non-informative priors allows you to do that, unlike the frequentist. I see. I guess I have another question regarding the benefit-risk analysis. So how is it different from the early phase trial where you find the toxicity? Well, benefit-risk analysis can be in many phase trials. It may not be in phase one, may not be in phase four. You can even do joint modeling in phase three. In fact, this can be applied at any time in the clinical trials. I know a lot of people think, oh, Bayesian is just dose finding. Or no, this is just real-world evidence and that type of thing. But no, any trial, this can be used. As long as you have learning and updates to that learning, you can use that. Even in planning phase three, for example, you have a bunch of phase two. What if you want to incorporate some of the learnings for phase two? Or you have a phase three in a rare disease and you have real-world evidence. You have something rare, for example, one of my companies, we worked on C3G at a palace and it was over and we didn’t know much data with that. But we had real-world evidence from three years ago. And you could use that as your prior for example. So for any phase of the trial, this could be useful.

Jocelyn: So on the regulation side of these clinical trials, do they look at these benefit-risk information a lot when they decide to approve it or not?

Carl: Well, yes, you can. It doesn’t necessarily even have to be a pre-hoc end point. Of course, it’s pre-hoc in terms of dose finding, that sort of thing. But of course, there’s a lot of post-hoc analysis that you do with a submission and that may strengthen your NDA. So you may want to conduct these analyses after. And that could help you as well.

Jocelyn: That’s really interesting. I don’t think I’ve ever done benefit-risk analysis before, but it sounds really interesting. So biostatistics is a very collaborative field and you must have crossed paths with a lot of individuals in the workplace that have a lot of experience on this. What are some skills or traits of individuals that make you think, oh, yes, that is the kind of person I want to work with?

Carl: I think one has to be very open-minded. Just because something may be correct statistically doesn’t mean that it’s correct clinically. And the same thing with something statistically significant may not be clinically significant. So as collaborators, we have to learn to understand the therapeutic area that we’re working with and really understand the clinician’s point of view and try to understand the medical part. Even though we’re not physicians, we don’t have to have a deep knowledge, but we have to at least understand what’s being discussed, what’s being talked about, and how that relates to what we do for our own analysis. Without having the context, it’s almost like a black box. And we don’t want to be thought of as somebody that is like a calculator, that we can just punch numbers into a machine and just come up with a result. There’s a lot that goes into that. And I think the best people to work with is somebody who listens and yet also puts their own idea forward. You also want to create your own idea and put that forward, but not necessarily be something consistent on that either, and adapt as you learn more. I think that’s the best person to work with.

Jocelyn: That makes sense. So for students or early career people who want to develop their career more on Biostats, do you have any advice for us?

Carl: Yes. So I think there’s a bunch of advice I would give. Number one, get an internship, which is important in any field. And I know sometimes it’s hard to get an internship, but I think the best experience is to have some type of work experience where you can actually collaborate with different functions. And this could even be your RA, research assistantship that you do in grad school, if you’re in a medical school or consulting center. I think you could have this collaborative environment with other functions outside of statistics and that’s the very basics that I think one would learn in industry rather than opening the textbook and doing questions or even experimental design. It’s not the same thing. So I think that’s where one would start. I started my first internship at the NIH before I went to Sanofi, and I thought that was very useful in a big facility like that. And internship in pharma would of course be ideal. Although you can’t control it, you could work on real data within the internship, that’d even be better. When I was at Sanofi for my internship, we couldn’t work with real data at the time, but I was able to attend an advisory committee meeting that was broadcast. And that was actually on Rimonabants, which was the drug that was pulled later by EMA when there was suicidal ideation that was caused by it. And I got to see how an advisory committee worked being broadcast in the cafeteria. It wasn’t a good day for Sanofi at the time, but at least I got to learn how this worked in the voting procedure and how each clinician would give their thoughts for or against the drug. Unfortunately, it didn’t work out for the company, but I learned that basic and it was very eye opening to me, even though I wasn’t involved, just sitting through there and being around the people discussing it is what’s needed. And I would say internship, if you could get one in person would be much better. That way, starting out, you really get to interact by Zoom. You can learn some things, but being in those conversations in the cafeteria, being in the office and hearing those conversations, especially starting out would be very beneficial. And the other thing that I would have to say is there’s the CRO, there’s a sponsor site and there’s the hospital site. They’re very different. They have their own purpose. The hospital medical research site, you’d be more like publishing papers. You wouldn’t necessarily be putting things into C-disc or SDTM format for an NDA. So the idea would be more to publish and there’s collaboration, but it’s more for publishing. Then you have the CRO where you’re kind of doing services as a sponsor, but sometimes outsource and you could also represent a very small sponsor, for example, that may not have a statistics function. You may be able to do that in some not so common instances, but you could. And there you would be doing a lot of task based things, maybe more programming in a way possibly, maybe more working for more clients instead of just one company, you’d be working for different ones. And that’s another thing that you would need to consider. Then the sponsor site, again, your goal is to get a drug on the market in different phases or to get more marketing for it, for example, health technology assessment or real world evidence. So I would say if you want to go on the sponsor site as a statistician, not a programmer, then definitely finish the PhD degree in school because that’s often a requirement by the job description, whether it’s really needed to do the job, that’s debatable. But in most cases, it’s a requirement to get your foot in the door in many cases. So I would recommend you doing that and doing an internship. If you’re more interested on the programming side, then maybe it’s better to get the master’s as quickly as you can and then start working on it. But keep in mind, statistical programming is not the same way or thing as software development. They’re very different. Statistical programming is more like getting the data and doing analysis into the type of format the FDA would like to see versus doing a bunch of algorithms in IT development for a web app, which would be more for software engineering. They’re very different types of jobs. So I would say at least get some experience knowing this is what you want to do if you’re into programming and that sort of thing. Then another thing is the hospital environment. If you want to be faculty again, then that’s getting the PhD and going through the tenure process and all of that. So again, I would say explore what you want to do and think in which direction and track. I would say internships are the best way to explore what you want while in school rather than going on to the market where it’s harder to make a U-turn.

Jocelyn: I see. Those are very good advice. Thank you. But I’m also wondering for people who had a taste in academia and want to go to industry or vice versa, would you say the skills or the experience are somewhat transferable or not really?

Carl: I would say they are. I know people that have started in academia and gone to industry. I think it’s a little bit harder going the other way. I go into industry rather than going back to academia simply because you wouldn’t be publishing as many methodology papers or having experience teaching. You could, there are some specialized groups in industry where you would do publications, but there’s not that many groups like that, but you could technically do that. But I think the skills in academia, especially if you are faculty, for example, in a research center, you’re working on getting grant money, that’s collaboration. That’s actually collaborating with other scientists, other functions and trying to do that, or even in the business school. So coming from academia to industry, yes, I know people have done that. Industry to academia is harder simply because you’re not going to do the tasks that academia would expect you to have or you’d be away from it longer. Let’s say you were doing a postdoc and you went to industry, you went back to academia, that would be harder simply because you would have been those years away from it. I would say from CRO to sponsor, absolutely, because you’re doing a lot of the tasks and you’re looking at many sponsors, and FDA to industry, absolutely. Again, similar to CRO, you’re reviewing dossiers from several companies and you’re providing feedback to the sponsors and you’re even communicating with them. You’re going to their meetings that the sponsors would have with the FDA. And to me, the FDA would be extremely valuable experience after your schooling because it would enable you to see how the regulatory environment works. And you have on your desk all these things to review and you get to see so many different things.

Jocelyn: I see. That’s very insightful. Thank you again for all the great advice, which I guess brings us to the last question we have for this podcast, which is, what is one question that you wish I would have asked and how would you have answered it?

Carl: Well, I wish you would have asked, how do you decide what you want? Coming off as an 18-year-old going to undergrad, which is a very difficult question for most people, even in some countries where somebody chooses a major and gets into a topic, sometimes they regret later and they end up going to the US, for example, and all of a sudden they’re doing a different track. So the question is how to know. And I think it’s a case of getting practical experience. I think learning something in school is one thing, being exposed to it outside of it is another thing, and there’s a transition to it. And I think that’s what someone needs to know. And I think in high school, it’s very difficult because you simply won’t have the exposure to that as much, even if you love something when you open the textbook. It’s not the same as actually doing it every day on the job. But I think one thing to ask is how to explore that immediately in undergrad and decide as soon as you can what track you want. I think it’s just getting exposed to as many things as possible and eventually narrowing it down. It’s that one thing that you would really like to do.

Jocelyn: I see. That is definitely very helpful. I do think I have the same problem, or I had the problem when I was in undergrad. Hopefully it works out for me in the future. Thank you, Carl, for being on the Biostatistics Podcast. And guys, I will link a YouTube video of Carl playing piano in the description. Maybe you can see how great he is in piano playing. Thank you for being with us.

Carl: Thank you so much.