Working with a pharmacometrician

Interview with Sree Kurup

My experience of working with pharmacometricians was very limited because I spent most of my time working more on the later phases. The contact with people in pharmacology in the later phases of the industry is not that big. That’s why I found this episode really helpful. If you don’t have a clue what happens in pharmacology, then stay tuned for this episode.

We discuss the following points:

  • What is a pharmacometrician?
  • What does the training of a pharmacometrician look like?
  • What is the role of a pharmacometrician in the development of drugs?
  • Besides statistics – what other functions do pharmacometricians work closely together with?
  • Where are the touchpoints of pharmacometricians and statisticians?
  • What is your experience of how both functions can work best together?
  • and more…

Listen to this episode and share this with your friends and colleagues!

Sree Kurup

Associate Director Pharmacometrics at Boehringer Ingelheim

She is a pharmacometrician at Boehringer Ingelheim. She started her training in pharmacy, and she had the opportunity to learn about pharmacometrics from the quantitative side of things. During her pharmacy background, although it was mainly clinical, she had the opportunity to work in a quantitative lab where she was able to use and learn about some of those skills to decide on compounds for development. She also had the opportunity to work as an intern at Boehringer Ingelheim and started her entire career in pharmacometrics.

Transcript:

Alexander: You’re listening to The Effective Statistician podcast, a weekly podcast with Alexander Schact and Benjamin Piske, designed to help you reach your potential, lead great sciences and serve patients without becoming overwhelmed by work. Today, I’m talking about pharmacometrics and of course I’m talking with a pharmacometrician about this topic. So stay tuned and learn how to best work with such a person. 

I personally have worked with pharmacometrician only very, very little bit in my career because I spent most of my time working more on the later phases and the contact with people in that space of the industry is not that big that’s why I found this is really really helpful and if you don’t have a clue what they do then stay tune for this episode. 

Subscribe to our Newsletter!

Do you want to boost your career as a statistician in the health sector? Our podcast helps you to achieve this by teaching you relevant knowledge about all the different aspects of becoming a more effective statistician.

I’m producing this podcast in association with PSI, a community dedicated to leading and  promoting the use of statistics within the healthcare industry for the benefit of patients. Head over to the PSI website at psiweb.org to learn more about all the different PSI activities who may help you become a PSI member today. 

Welcome to another episode of The Effective Statistician and today I have a person from a related field. Hi Sree, how are you doing? 

Sree: Hi, Alexander, very good. How are you? 

Alexander: Very good. Thanks so much. It’s nice to have you on the call especially as you are a listener of the podcast even though you are a pharmacometrician and not a statistician but there’s a lot of similar challenges and a lot of similar things that we both kind of work with as quantitative people, that’s really, really great. Before we go into the content, maybe you can explain a little bit how you got into the job where you are now and what motivated you to go into this direction. 

Sree: Yes. I’m a pharmacometrician at BI, working here for about five years. I started out my training in pharmacy and I actually got the opportunity to learn about pharmacometrics from the quantitative side of things. During my pharmacy background, although I was mainly in clinical, I also got the opportunity to work in a quantitative lab where I was able to use and learn about some of those skills to make a decision about compounds for development. And I also had the opportunity to work as an intern at BI and so I actually started out my entire career in pharmacometrics at the intersection between pharmacometrics and statistics where I worked together with another statistics intern to work on MCP model actual use, so those exposure response modeling. That’s how I got into pharmacometrics and learning about the quantitative side of things and how we can integrate the drug, disease, and trial characteristics to basically understand the system mathematically and then use that to make decisions about drug development, that was very appealing to me so I’m very passionate about using quantitative approaches to make better decisions and drug development.

Alexander: During your initial time at BI as an intern, was there a specific episode or specific story where you were realizing, ‘this is really cool, I want to continue with that? 

Sree: Yeah. I think, during the internship I worked with a person from statistics, like other interns in statistics and we were working on the MCP methodology where we want to integrate this multiple comparison testing with the modeling approach to select the dose for phase 3 setting and this was all based on dummy data, so it wasn’t really high stakes or anything but just being able to account for the uncertainty in models and then be able to use that as a way to make a decision about which dose to take forward into the development.

Alexander: That was cool. That’s a pretty nice story. MCP models are truly not some kind of completely straightforward approach and working on these things is really, really nice. When you look back into your training as a pharmacometrician, how did that actually look like what kind of areas do you study? Because in statistics, I studied mathematics actually, we did all kinds of different models and things like this but I didn’t really learn a lot about biology or anything like this, so how is it in pharmacometrics? 

Sree: Yes, basically. I started out in pharmacy so I had really no training at all in quantitative sciences. So I had good training in Pharmacology and understanding the relationship between dose and like a concentration of drug in the body and how that leads to downstream activation of some biomarkers and the physiology, pharmacology side of things. And then I have the opportunity to basically learn some quantitative skills, programming skills, modeling skills, when I had to apply the knowledge that I have and kind of accumulate the knowledge in a mathematical model to then be able to say, ‘okay, what if we use a different dose? What if we use two different regimen or route of administration, and how can we use the knowledge that we have so far about the compound, including the biology, physiology, pharmacology, to make some kind of simulation or answer a hypothetical question to say, ‘Okay, what will the exposure look like? What would the response look like in this scenario? And how would that impact an outcome that we were interested in or a research question? So, basically, starting with the question in mind like, ‘What is the research question that you have in mind’,using a model to basically, accumulate all the knowledge that you have, and then apply that for some decision making. 

Alexander: What are these typical kinds of questions that you would come across as a pharmacometrician?

Sree: Yeah, so it can be from discovery until approval and post marketing and beyond but you would want to know things like what’s the relationship between dose and exposure because that relates to what should be the dose to give to get to a to be able to target, the exposure to be able to reach a certain level of response, so dose selection.

Alexander: In terms of exposure, that means how much active ingredient we have in blood, for example, that would be exposure. 

Sree: Yeah, that’s exactly right.  

Alexander: Okay. Is it always about what’s in  the blood or can it be also other ways of exposure? If I’m thinking, for example, for something that works in the brain or maybe it works in the skin or things like this, how does that work then in these kinds of areas?

Sree: Yeah, exactly. So usually what you can measure is what’s in the blood but of course, what’s really related to the effect and downstream activation of the pharmacology and then eventually to the efficacy is what’s actually in the actual site of action. So first, you think about the concentration of drug or the dose that you give and then the exposure in the blood and then from the blood it distributes into the tissues and then it activates some receptor or pathway and then leads to a response downstream. Well the principle of pharmacokinetics, so to understand the relationship between dose and exposure, you can have an equilibrium between the blood and the tissues usually so you can describe that mathematically with a model. So if you think about the drug, the dose first and then characterize the body as sort of like a compartment and then you have the drug coming into the compartment and then distributing and then being eliminated, absorbed, distributed throughout the body and metabolized and so we can capture all of that with the model. And if you’re really interested in specific tissue because you think that’s directly related to some question you want to answer then you can also use physiologically based models. So the typical models that we work with in PopPK or population pharmacokinetics are mainly empirical so it’s not the human body, it’s can’t be represented, so it’s very simplistic representation of the human body, but you can also have like physiologically based models where it’s like actual tissue volumes and blood flows and organ function things like that. So, this is a separate kind of modeling approach where you look more into the physiology and  the blood flow and things like that to get an understanding of the tissue concentration as well but typically we can answer those questions with a simpler model. 

Alexander: If you look into exposure and I think of, ’Okay, I take a drug and then I measure the amount of ingredient of what in the blood’, I think there’s a lot of different things you can measure in it, probably how fast it increases, how much you have over a period of time, how fast it goes down, how it metabolites, how it leaves the body again, all these kind of different things. That’s all PK, yeah?

Sree: Yeah, you’re exactly right. So, basically, when you think about pharmacokinetics, you want to look at the time course of the exposure in the body. So, if you think about a subject in a trial, getting a dose of a compound, then you want to see how the pharmacokinetics look like, then you will want to take the blood concentration or the concentration of drug in the body over time. So basically, you would collect multiple samples and then look at the data, basically and then you can see how the drug absorbs the shape of the curve basically and so there’s different characteristics like the absorption rates, how fast it gets absorbed, if it’s an oral compound or a different type of administration route. You can also see how fast it declines like you said. How long does it stay in the body? So like a half-life concept. So all of those things can change depending on the dose, the dose interval, the route of administration and there could be variability even between subjects. So, depending on how it metabolizes either through the kidneys, liver or catabolism, there can be differences between patients that could impact the relationship between how much is in the body and how fast it gets absorbed or eliminated or metabolized. And if you have a different organ function than that can also possibly impact the resulting exposure. So there’s like variability between patients but also just random variability as well. So we tried to characterize that relationship using a structural model, looking at a typical profile over time but also stochastic elements. So like just mixed effects modeling where we want to include random effects. Just on a parameter, we would include some level of variability and make an assumption about how that parameter is distributed. So typically, for PK, it’s not log normally distributed, so things like clearance and volume. We try to get an understanding, like a quantitative understanding, of the main PK parameters that would help us learn about how long it stays in the body essentially. 

Alexander: Yeah. And if we now have for example something that goes too fast or that goes not higher up or decreases too fast again, how can we influence that? 

Sree: It’s basically the drug that you have and it’s the characteristic of the compound, right? And so, you can engineer the molecule in such a way to make it have different properties like a half life extension, for example, of an antibody but usually that’s what you’re working with. Like in the preclinical space, you have some information and so you can select compounds that give you the PK characteristics early on in the discovery phase. And then once they get into man you learn about that and then you see, ‘Okay, how often should I give this drug, what route of administration, how much of it do I need and how often do I need to give it to be able to reach a clinically efficacious, kind of exposure’. 

Alexander: Can we then also have to go back to the bench and have different kinds of formulations? Is that also part of this area? So different mixtures or instead of having a tablet, having some capsule or other ways to formulate that so that it better meets the demands of the patients. 

Sree: Yeah.

Alexander: Okay, interesting. You mention there’s a lot of collaboration with statistics, how does that actually look like? Because both are quantitative scientists and how do you divide up the work and where’s the overlap?

Sree: Yes. As a pharmacometrician, you’re typically also sitting in those drug development teams together with these statisticians, of course but also the clinical lead, clinical pharmacology, biomarker colleagues and so forth and you’re all in this team and trying to develop a compound, right? And so when you think about designing a study, for example, then defining the right research questions together and thinking about how to best select the dosing regimen or how to design the study to get the most information out of it. And so I would say thinking about the just, for example, like designing a Phase 2B study and thinking about what dosage should be evaluated for the optimal dosing to be able to learn about dosing in a Phase 3 setting, you would work together to think about, basically the design, what exposure could be driving the response or what exposure metric, when is PD study state reached, when is PK study state reached, which population or disease course is being studied and working together to optimized the trial design together. And I would say that it’s really starting with the question that we have in mind, defining that together and like one place that’s kind of there’s a lot of synergies, exposure response and dose response. And designing a dose finding study or thinking about which dose to select into phase 3, for example, in a late-stage program. If you want to use MCP mod, for example, you might want to think about which shapes should I pre-select into my mCP like the design stage. And there we can use the pharmacometrician knowledge in the sense that you would have accumulated knowledge about the compound that the disease, the relationship between the drug and response and things like that. From preclinical, basically, prior knowledge, so it’s kind of Bayesian in a sense. So working together to kind of look into that space like they could the shape of the exposure response, how to best design the study and also interpret the study, analyze the study. Sometimes the dose response is good enough, sometimes we need more complex models or you want to take into account different aspects and different covariates and things like that. So because with the dose response, you would just assume the same average response with the same dose, whereas that  might not always be the case and exposure, what’s really in the end related to how much is at the site of action and in the end leading to a clinical outcome. So if you have different formulations or different routes of administration in the end and your previous study didn’t really look into that, let’s say you had an IV administration in your phase 2B or phase 2A, now the team wants to go for a Sub-Q development, without really understanding the relationship between dose and exposure, you can’t design that appropriately. 

Alexander: I love how you describe it. It’s really a collaborative effort. It’s not so much, ‘that’s my sandbox, that’s your sandbox’. But it’s, we need to work across the team, with all the different people and give all our background into it so that we make the best decisions in terms of design and analysis of these studies to move forward and thinking from having the end in mind, in terms of, how will approval look like? How will phase 3 look like? How do we need to have Phase 2? I really love this approach, highlighting the collaborative efforts here. In terms of working together in such a collaborative space, what are your experiences? How to make that best work?

Sree: Yeah, I think, one of the things that we have to keep in mind is really thinking about the first, the question and then the data that we would have to answer that question. And then think about the approach that you would take. Sometimes the approach requires let’s say more assumptions around the model structure or how complex things should be and what those parameters are informed by and things like that. So, trusting each other’s knowledge, I would say, quantifying the uncertainty if you don’t have, you know, as much knowledge about a parameter.

Alexander: How do you gain this trust with a statistician for example, that you can trust his knowledge or her language?

Sree: So I would think starting off with just some common ground, right. So I think obviously even pharmacometrics that started out from statistics because it’s borrowed from mixed effects modeling, right. So we have this common understanding of mixed effects models. So we have, you know, defining those terminologies and making sure we understand what we’re talking about even, sometimes it’s not exactly the same terminology. So, I would say just talking to each other and learning about each other’s discipline and seeing where you have common ground, usually there really is. The pharmacometrics are borrowed from statistics, and the other way around. 

Alexander: I think that’s important, being curious about the other person, where the other person is coming from, understanding the language or vocabulary and then kind of listening to what is the other person’s expertise? Why are they kind of thinking, ‘Okay, this model would make sense here. These assumptions would make sense here. These assumptions wouldn’t make sense here’, that helps to see where the other person is coming from. Do you have any story of this kind of collaboration between you and statisticians that work really, really nicely?

Sree: Yeah. Actually even now, we’re working very closely together to better assess the benefit risk of a compound. And so looking at, you know, we have many different trials, like 20 trials, or so. And we want to understand the relationship between dose and let’s say, safety. Like is there a dose-related concern of safety and actually, in that case, we can also think about exposure and safety, because there are so many differences in the trial design types of studies that were done along routes of administration, things like that, that if you look at just dose, it may be misleading. And so in this case, the statistician actually used exposure as the explanatory variable looking at the relationship between dose and response. And so in this case, I simulated the individual subjects’ exposure values for that study and then we defined an exposure metric together and then assimilated that individual subject PK value and then gave that to the statistician to then be able to plug that into an exposure response model. In that case, they actually did that themselves. So that was a nice example, where I think the statistician also learned a little bit more about exposure metrics, the limitations of certain metrics versus others and we shouldn’t be pulling everything together, maybe should look at certain things. And then also looking at, let’s say the impact of different covariates on the relationship of dosing safety, for example, because maybe not all of the different diseases that we included in that model are appropriate to pull together and things like that. So I think that was a nice example where we worked very closely to be able to answer that question that came up from a regulatory agency in a very short timeframe.

Alexander: Cool. That sounds really really interesting because you’re clearly measuring plasma levels is quite a lot of effort and simulating these based on limited data that you have. It’s quite a neat way to understand, you know, the exposure to side effects the process may look like and to then better understand what kind of actions to take because certain patients may be at higher risk and things like this. Very, very cool. Thanks so much for this discussion. We talked a lot about what pharmacometricians do, what their background is and how statisticians and pharmacometricians work closely together and what good collaboration is between the two functions. Is there any kind of final tip you would have for a statistician to work effectively with a pharmacometrician. 

Sree: Yes, I mean, I would say we talk to each other. 

Alexander: This is really an obvious one on the first glance but I know that especially when we all work remotely, we very often meet in meetings rather than on a one-to-one basis. Make time to meet people one to one. And after the pandemic, maybe even face-to-face. 

Sree: Yeah. Definitely like just learning from each other, talking, understanding the same language, sharing the same language and then really collaborating on the questions that you want to answer together and then digging about the approach going from there. So, just go talk to each other and maybe something happens. 

Alexander: Thanks so much for this great discussion. Talk to you soon. Surely again at another point. 

Sree: Thank you. Thanks for having me. Alexander: This show was created in association with PSI. Thanks to Reine and Casey, who helped with the show in the background and thank you for listening. Reach your potential, lead great sciences and serve patients. Just be an effective statistician.

Never miss an episode of The Effective Statistician

Join hundreds of your peers and subscribe to get our latest updates by email!

Get the shownotes of our podcast episodes plus tips and tricks to increase your impact at work to boost your career!

We won't send you spam. Unsubscribe at any time. Powered by ConvertKit