Add-On to All-In: Building a Real AI Strategy in Clinical Research

Jimmy Bechtel sits down with Dr. Anastasia Christianson, former senior innovation leader at Pfizer and Johnson & Johnson, to talk about what a real AI strategy actually looks like in clinical research. They dig into why isolated pilots fall short, how to think about AI as a process change rather than a tool add-on, and where the biggest missed opportunities are right now. Dr. Christianson also shares her take on navigating evolving regulations and paints a picture of where clinical trials are headed over the next three to five years, including a future where no trial launches without first being simulated end to end. 

Transcript

Jimmy Bechtel 0:00

Thank you for joining us as we explore the latest trends. Insights and innovations shaping clinical research today. I'm Jimmy Bechtel, the Chief Site Success Officer with SCRS, and today I'm joined by Dr. Anastasia Christianson, who is a former senior innovation leader at companies like Pfizer and Johnson and Johnson, and now is advising some life sciences leaders as a strategic advisor. We're here actually to talk a little bit about AI and what's going on in this space and, and how we can look at AI strategically. But before we get into that topic, Dr. Christensen, it'd be great to learn a little bit more about you and your background, if you would please.

Anastasia Christianson 0:41

yeah, absolutely. First of all, thank you very much for inviting me to join this conversation. I've been listening to your podcast. I'm really excited to be part of it. So I'm currently serving as the managing principal at EPAM, life Science Consulting, and also executive and residence at Columbia Tech Ventures. This is after 30 years in the pharmaceutical industry in four major pharma companies across the entire drug discovery and development process. Start starting in discovery. leading drug projects, then moving to clinical development and enterprise functions. Most recently I was senior vice president and global head of AI data and analytics at Pfizer. And prior to that, I held executive roles, as you mentioned, at Johnson Johnson, Bristol, Myers Squibb, and AstraZeneca, where I drove. Data excellence, digital transformation and innovation across r and d and, and especially in clinical development. Very passionate about this topic about using ai, which maybe seems new in the past few years, but we've been using it for many, many years. And the transformative power really of data and AI in healthcare.

Jimmy Bechtel 1:46

It's absolutely a really exciting opportunity and a really cool forefront for us to be on. So, it's great to have someone, with your background, Dr. Christiansen here, to. Talk with us a little more about AI and, and what it means in the life sciences and like you had mentioned. You've been working at the forefront of innovation at, some of our partners, major pharmaceutical organizations. Let's start off by talking and getting your thoughts around what a forward thinking AI strategy actually looks like in practice and where organizations might focus first and, and remembering as well. Running the gamut here from our audience, so talking maybe to sites, but also talking to maybe some other, industry based partners as well.

Anastasia Christianson 2:29

it's really important for organizations to treat AI as part of how the business runs, not think of a separate AI strategy, but rather, think about. What does the business need? What are the biggest bottlenecks? And think about the business end to end, right, from not just little, you know, bot, small bottlenecks in specific areas, but solve those, but within the context of the bigger end-to-end. process that you're running. So isolated pilots are okay to start with specific use cases, but in the end, you need to be grounded in clear business value in, the entire process and in how you're gonna scale a small use case to actually fit within the broader process. So with that in mind then, the quoting core ingredients of a strategy is a business strategy first. for a world that uses ai, based in business objectives, priorities. And then back solving for where AI can uniquely help. so not very unlike what we were doing with digital transformation or automation or digital health, where, sure you can do automation for the sake of automation, but what about the problems you're trying to solve? Start there, apply automation and, the most important part is to start with a problem at hand and then look at. Is AI the best approach, the best solution? Because you wouldn't take a sledgehammer to put a nail in the wall, right? 'cause you might actually break the wall. So, you know, sometimes you gotta think about is AI the right solution for this problem? And if not, what are other. solutions that you can use that might be more AP applicable. and so aside from that, then explicit principles and guardrails. So your strategy should have some guardrails that define a core set of principles of, for your business, that your business will abide by, with, you know, from the perspective of safety, transparency, human oversight. IP protection, et cetera. we should really keep, maintain a portfolio of business problems that you are solving, and how you're solving in them. And then think about your overall operating model. So in practice, when you have these AI solutions that you're, your, developing, how's that gonna fit within your operating model? Or rather, how will it change your operating model? You don't really wanna. Shoehorn it into your current operating model. 'cause that defeats the purpose. So you really need to think about the, the big picture where AI fits in and any solution that you develop. How does that modify your current operating model and work on that?

Jimmy Bechtel 5:08

So many interesting points, Dr. Christensen. Starting with, this concept of governance and making sure that you have the right institutional guardrails in place and you've of drawn the lines in the sand around what is appropriate use and where and when we should apply various different aspects of AI to our organization. So I think that's, really essential. But also what I found really interesting about your response is this, this concept of working it into what you're already doing and trying to find ways in which we can apply AI into our current, operations. that approach is so different than what we're really used to in research. You look back on things like eSource and EDC and risk-based monitoring and all these concepts that have come to fruition over the last few years, and it's been this new sort of tack on or a new thing that we're just adopting and we're being asked like, Hey, we're gonna go about doing it this way, or We're gonna add this service into the clinical trial. And that's not what we're talking about here with ai and it, it's an incorporation into the vast aspects of what we do in clinical trial execution. So I think it's really important and a point well raised when we think about it, we approach it as not just. This tack on or this new thing that we're being asked to use, but think about how can it augment or bring to life some of what I'm already doing in a new or improved way?

Anastasia Christianson 6:34

Yeah. Or change what you're doing. Right. Because sometimes we try to just fit it into our process, but in reality, you probably need to change the process and the roles, and you should look at that first and foremost. Design that as part of rolling out your AI solution. Otherwise, it becomes a matter of, well, some people might adopt, it might not. They might, you know, dig their heels and not wanna adopt it. But if you're changing the process along with giving them a better approach, a better tool that incorporates ai, then they're more likely to see the value and they, in a way, you can't go back because you're approaching it in a different way.

Jimmy Bechtel 7:11

That's an another excellent point. I couldn't agree more. Dr. Christensen, we see many companies are still sort of playing it safe with AI and they're dabbling or they're doing, you know, kind of the basic stuff that everybody does and they're this, that, or the other, and chat GPT and things like that. But what are, from your perspective, some of the biggest missed opportunities you see in clinically applied artificial intelligence right now?

Anastasia Christianson 7:37

Yeah, I'm gonna start with a point I raised a little bit earlier, which is most people are looking at one step in the process at a time, and we apply AI to improve that one step, and then we move to the next one, and then we try to figure out if that's gonna help us across the board. So looking at the process holistically. So for example, if you look at clinical trials, right? 50% of trials. Probably more, but that's a good number to work with, actually fail to meet their endpoints. And then you look at why do they fail to meet their endpoints? And you come up with, well, partly it's trial design, partly it's patient recruitment. recruitment doesn't come alone with just recruiting. It's also retention and, outcomes for the patients and monitoring and keeping them and keeping the patients engaged. And now you look at not just improving. Patient recruitment as one step, but you recruit them and then you don't keep 'em in the trial. That doesn't help you 'cause you are still gonna end up with not meeting your endpoint. So if you're gonna recruit the patients and you're gonna be looking at ways to improve on that, you need to talk about the entirety of engaging patients from the beginning before they get recruited to actually recruitment to, engaging them during the trial, allowing them to answer questions. be monitored, be able to a, to ask questions and get a quick response. Help them, sign up for, a testing, a test that they need to take a, help them with, scheduling and so on. So if you. Look at the entire process and think about engaging the patient not only in the recruitment, but throughout the trial. Now it's a different solution that you're looking at and a different way that you are using ai. And you can use AI for every one of these steps, and it's not a different solution for each of these. It's actually one solution that engages the patients from start to finish.

Jimmy Bechtel 9:26

Which again, Dr. Christiansen is, is a, a relatively novel approach and something we have to kind of get ourselves into a quote unquote new mindset around. Because for so long as an industry, we've operated with these singular siloed solutions. One, you know, we have things that help with recruitment. We have things that help with retention to you. You know, reference the, the example that you gave, but to your point, artificial intelligence AI can be applied across that spectrum from start to finish with patient engagement and make the entirety of that opportunity. Something a little bit more seamless or effective and efficient, for them, which, is to your point right, an an excellent opportunity. An excellent application for, for these sets of tools that we have, available at our fingertips.

Anastasia Christianson 10:12

Absolutely. And if I could add one more missed opportunity is. Simulating the trial before you actually start recruiting patients and before you've made decisions on sites and so on, you can simulate the trial with the data you have using real world data, using the exclusion inclusion criteria that you have different geographies, you know, synthetic virtual controls and so on. So that you know where the bottlenecks are going to be and you're looking for to resolve the bottlenecks. So when you are, you actually start the trial, you have less surprises. I think that's another miss that we're starting to, to solve for right now. So I, I don't mean to say like nobody's doing it, but we're not doing enough of it. yet

Jimmy Bechtel 10:57

I would agree that's probably a, a really great opportunity, for us to, yeah, really identify some of the challenges and some of the bottlenecks and some of the issues that we might encounter when we go to actually enroll live patients, in the clinical trial. So, a couple of great examples. There are probably ones that we could spend, a lot of time talking through and teasing out the nuance. but, we'll, we'll move on for the sake of, for the sake of conversation here. Dr. Christiansen, what's your advice then for those that are navigating the regulatory landscape around AI and other digital innovations? as we've seen these guidelines and, and really the capabilities of these tools are constantly evolving, it seems like we, something comes out and we're able to understand what its purpose is and people start to use it. But then boom, here's a new use for AI and here's a new application for it and it can do this now. So how do we, again, how do we navigate that landscape? most appropriately.

Anastasia Christianson 11:56

The regulations are evolving, but not randomly. They're evolving because AI is evolving every other day. Yeah. AI has evolved a new tool. It's a new capability, and it's really exciting. But it also challenges our regulators to kind of stay on top of it. So I would say a couple of things. One is engage the regulators early and often. So stay on top of changing regulations and engage the regulators in conversation, especially as you're using ai. To make sure that they have all the information they need, that they're comfortable with how you're using ai, that they can ask questions before you get to the point of, of filing. And the other part I would say is, keep track of the AI tools that you're using. So you're keeping track of the regulations, then the changes in regulations, you and talking with the regulators, early and often. you're also keeping track of the tools that you are using that are using AI so that you have. Clear classification of your ai, classification in terms of risk classification and what it's actually doing. You have standing governance within your organization on any AI tools, either that you're building or that you're purchasing from vendors. So you're clear on what, what that AI tool is doing and how it's being used and how you're using it. A regulatory intelligence function that feeds into design and validation. So you're making sure that you are constantly thinking about the regulations and keeping up with them as a trial is ongoing because you may have been fully aware of the regulations when you started the trial, but a year into the trial, something changed and now you need to. evaluate and see what you need to change so you can evolve quickly as the regulations are are evolving.

Jimmy Bechtel 13:44

Basically what I'm hearing is making sure that you're consistently. Aware of what those regulations are and what systems you are using as they apply to those regulations so that when changes inevitably come, you're able to move quickly and be nimble about those changes and those enhancements really, and, and understand more effectively how they're going to, change exactly what you're doing right. A lot of those things, as you alluded to. Those changes will modify potentially how you are using AI in a specific application. So being on top of that essentially allows you to be, nimble with it.

Anastasia Christianson 14:24

Absolutely.

Jimmy Bechtel 14:25

So Dr. Christensen then to con start to conclude our conversation here, one, one sort of final question. How do you then believe AI in clinical trials will ultimately evolve over the next maybe three to five years? And getting at our sort of theme du jour here, how can organizations continue to be prepared for that work?

Anastasia Christianson 14:47

Yeah, in a way, almost an impossible one to answer, being that AI is continually evolving and the, the technology is evolving. So I'd like to think that within three to five years, no trial will be initiated without first simulating it and optimizing it. In silico, if you will, in computer before you actually, finalize the protocol and start enrolling patients. It's not farfetched, it's not that far away, and maybe we will achieve it within three to five years. If we think of the aerospace, engineering, domain, for example, no plane goes. on its first flight without first doing a series of flight simulations. So it's not unreasonable that we would do that. And I just use that as one example. There are other, areas that, do the same where they, simulate before they actually go to market or before they take flight or what have you. I think that's what we need to do with our clinical trials. We have enough real world data, we have enough operations data. We can actually simulate the process and we can learn and continue to, to improve the simulation. So that's one way that clinical trials, I believe, will, change maybe more than five years, maybe within the five years. AI will move from having helpful tools around the edges to actually accelerating and enhancing. Big steps in the process. AI will be embedded in decision layers in shaping the design and shaping the conduct. when we start looking across the board and start, stitching together some of these solutions, we'll be ready for. full on simulation of the trial. So we'll have smarter, more adaptive trials that are simulated or simulation driven and, adaptive designs that become routine. And you can hopefully half the time that it takes, to run a trial. but at the very least you won't have this. Stop and go. you won't have the dropout of, patients. You'll be able to have a, a smoother ride.

Jimmy Bechtel 16:46

I think that's what really excites me about what you just said is that we're able to. First and foremost, not only bring medicines to patients who need them faster, I think that's obviously the ultimate goal and the end result of what you had just talked about. But along the journey that we have, which is very bumpy right now and challenging. We're able to use AI again, removing it, moving from the periphery into the embedded model of clinical trial delivery, and really accelerate the work that we do as an industry and, and not only accelerate it, but also, you know, sort of. then removing all of, or most of the challenges and the barriers that cause those delays and cause things to be, to be challenging. So again, really awesome perspective and I think a really great place for us to end our conversation on that sort of, optimistic and and perspective. High note, as we move into this, Dr. Christensen, we could talk for days on this, I think, and I'm really excited, for, for more opportunities to engage with you on this. But thank you for your time today and thank you for engaging, on this topic and, and providing some, some introductory light, in, in the industry to, where we are headed with artificial intelligence.

Anastasia Christianson 18:02

Absolutely my pleasure. And I'd love to come back when you're, when you're ready. That would be fantastic. And let's keep in mind that the, AI is continually evolving and we're evolving with it, or our processes are evolving with it. And I think at some point we're gonna be able to model disease model patients and, go much, much faster and more importantly, get better outcomes on the trials.

Jimmy Bechtel 18:24

That's exactly right. Well, thank you again for your time today, and for those that are listening, I hope you check out other great site focused resources made available to our entire community on our website, my scrs.org, including other podcasts, webcasts, and other AI related, events and content made available for our audience. Thanks for listening, for tuning in, and until next time.