From Snapshots to Continuous Evidence: How We Learn

Clinical trials provide critical answers, but they're only part of the story. In this episode of The Health Pulse, Aaron Berger and Dr. Massoud Toussi, real-world evidence executives at United BioSource (UBC), explore how the life sciences and health care ecosystem is moving beyond isolated snapshots toward continuous evidence generation. They discuss how organizations are building a more complete picture of treatment effectiveness, safety and patient outcomes in the real world. Electronic health records, connected devices, patient-centered research and global evidence networks are making that possible.

Making this vision a reality requires data quality, interoperability, regulatory alignment and trusted evidence generation. AI is playing a growing role in research and analysis. It is creating new opportunities to accelerate evidence generation and uncover insights at a scale previously impossible. 

Ultimately, they share an optimistic vision for the future: one where real-world evidence, AI and modern data ecosystems help us learn faster, answer more questions and accelerate scientific discovery for patients who need new therapies most. 

Creators and Guests

Host

Alex Maiersperger

SAS

Guest

Aaron Berger

UBC

Guest

Massoud Toussi

UBC

What is The Health Pulse?

How can data, AI and advanced analytics accelerate health innovation? Which new technologies hold the most promise? What are the biggest roadblocks to progress? How can we solve endemic problems?

Join us for The Health Pulse podcast series as we explore fresh perspectives on digital transformation in health care and life sciences. With a special guest expert on each episode*, we’ll tackle the most pressing issues affecting the delivery of health care and therapies worldwide.

All presentations represent the opinions of the presenter and do not represent the position or the opinion of SAS.

MASSOUD TOUSSI: Let me tell you about a personal experience. Last year, I had a sabbatical, and I started working on an AI-- for example, an AI system which would allow building protocols, statistical analysis plans, et cetera.

And once I got finally the first version of the protocol, which was generated by the system, I couldn't sleep at all, because the main thing I was thinking about was, what are we all going to become? The AI is-- this AI generates something that I wasn't able to generate in terms of quality, in terms of extensiveness, et cetera, in terms of time, efficiency, lack of errors, lack of editorial errors. Whatever you could think about it, I think that I wouldn't be able to generate what this tool generates.

[MUSIC PLAYING]

ALEX MAIERSPERGER: You've heard of humans in the loop. Can taking humans out of the loop improve outcomes? Today, we'll hear from global, real-world evidence leaders at UBC who are making sure we design studies around questions and not just the data. The better the question, the better the answer. And in this case, that means better, safer, faster drugs.

We're going to talk about real-world evidence. We're going to talk about data, AI, and analytics, and the role they play in effective real-world evidence production, management, deployment, and how you do this across local teams and global teams. What does-- Aaron or Massoud, we'll start-- either of you that want to chime in on the first one. What does good data actually mean in practice? You hear this about AI-ready data, or just good data. You need to start as the foundation. What does that actually mean in your world?

MASSOUD TOUSSI: It's an interesting question and very timely, because just a couple of months ago, EMA issued their data quality framework. There were also other frameworks in the past, published by ISPOR or FDA. So generally, when we're talking about data, we think about a number of dimensions or criteria, the most important of which is fitness for purpose. The data needs to answer the question that we want to answer.

But if we make abstraction from that question, which is not always there when we are evaluating a database, then there are a number of other dimensions that come to mind. The completeness, for example, the data. How complete is the data? Which means how much missing information we have in that data. How valid or accurate is that data? Which means how many or how much errors we have in that data.

How timely is the data? In a lot of situations, for example, there is a data lag. You cannot use that for a lot-- for your needs because it doesn't answer your expectations in terms of timelines. And also, there is coverage. Sometimes you need your data to be representative of the population that you want to study. So if the coverage is not good, you cannot actually use that data.

ALEX MAIERSPERGER: Sounds like starting with the question in mind, that fit for purpose of, we've got to actually have the data that's going to answer the question were asking. So a lot of it comes down to that question. Is that right, Aaron?

AARON BERGER: Yeah, and the only thing I would add to that is that even though we're in this age of modernization and disruption with new technologies-- we're going to talk about AI, of course-- the same themes are universally and still consistently applicable when it comes to data, the concept of traceability and provenance. And then when it comes to the outputs, are they reproducible?

These were-- this was all true as being important themes before we started talking about AI. It's been very true as part of the real-world evidence discussion for many years, and it's still true today.

ALEX MAIERSPERGER: Evidence generation touches everything. And so trials, commercialization, patient access. Massoud, where is it still failing today, and where is it finally starting to deliver real-world impact?

MASSOUD TOUSSI: Yeah, I think the evidence generation in the past two decades has evolved a lot. If you want to consider the entire spectrum of evidence, which includes also the clinical trials, I would say the evidence generation is still episodic. So it's like the difference between a camera and-- a camera which takes movies or films and a camera which takes just pictures. And we are still in that episodic view of evidence, because we have snapshots of the evidence, not continuous evidence.

And I think we are now-- there are a few examples, actually, in the world which have been conducted as MVPs that show that actually, it's possible to move toward that direction of continuous surveillance evidence generation and having real-time evidence generated using the information which comes from the point of care, and then generating real-time evidence, which also serves the point of care, which is like a feedback loop. And there are a few examples about that.

AARON BERGER: Yeah. And I would recast the question just slightly and say, I don't think of real-world evidence as failing, often. I do think there are some of us out there in the evidence generation community who would say that we would have liked to see more applications and more consistent applications of real-world evidence.

We have examples of label expansion and new indications being approved with real-world evidence, post-marketing requirements being addressed with real-world evidence, safety questions. But perhaps we'd like to see more of that, and more acceptance and more adoption in that space, and maybe just a faster climb up the adoption curve there.

ALEX MAIERSPERGER: I really like the analogy of the flash camera and the video camera. And obviously, we want that for all of us. Going to the doctor, you want the video record of your life versus the one single snapshot. You want to be able to say, here's what's going on and why.

Massoud, you've compared drug comparisons to car miles per gallons, the sticker versus the real-world driving. And so there's obviously a difference there of, hey, my car is not getting what's on the sticker in the real world. And obviously, you're going up different hills, and you've got different stuff in the car. If medicines had a real-world label, what would be on it?

MASSOUD TOUSSI: Yeah. I would actually-- in that dashboard, I would put, first and foremost, the real-world effectiveness of the drugs, how they are-- basically, there are a few questions that we are asking ourselves for all drugs. So if I want to list them and put them on my dashboard, that would be, what is the real-world effectiveness of drugs? How many people are actually using that drug? We would like to know that. What is the characteristic of those people? And then what is the treatment pattern? How this drug is used in real life.

And then what are the safety concerns? So safety concerns, it can be also adverse events or adverse effects of the product, but also about the risk factors, which predispose people to have those adverse effects. So the risk factors of those consequences.

And then finally, there are also a number of things, such as, for example, if the drug is being used according to their label. So it's about off-label use. And whether or not, for example, if there are any kind of risk minimization measures or REMS programs, if they are followed or not. So these are the things that I would put on that dashboard.

AARON BERGER: Yeah, I really like that vision, Massoud. And then it makes me think about all the things that would have to be put in place from an infrastructure standpoint and data acquisition standpoint in order to bring that vision to reality on-- that's kind of the other end of the pipe, the dissemination. What do we learn from setting up real-world, real-time registries, capturing patients' EMR data, capturing their claims data, capturing their experience data, quality of life, PRO?

Putting that all together and then turning it into the outputs that you've talked about will require that real-world, real-time type of study to be set up in advance. And that's the part that gets me excited, because that's exactly what we do in designing direct-to-patient studies that capture that type of real-world data without using health care providers all the time as the intermediary between the data and ultimately our collection efforts.

ALEX MAIERSPERGER: I know for a long time the industry talked about population health and the way you would look at a broad population and be able to intervene in certain instances. I think now you hear a lot more about personalization. And so that sort of answer just really reminded me of the personalization of, right now if you have a population on a drug, you're sort of lumped into that population. You say, we're all diabetics or we're this condition.

Versus the personalization of saying, somebody else my age who also has this condition, who also has x other things going on in their life, this is how the drug will interact with me individually. And so I'd love to hear the generation from population health to real personalization, because all of us individually want to be treated as individuals.

Aaron, real-world evidence sits between regulators, payers, pharma, providers, and patients. And so you've got a lot of different stakeholders in the mix, and they don't all want the same thing, and they don't all sort of-- in the hierarchy of values, it might be speed over quality for some, or it might be quality over x for some. Where do you see the biggest tension today?

AARON BERGER: The biggest tension. Well, I guess one area of tension-- I don't know if it's the biggest one, but I guess it goes back to my first-- one of my earlier responses on, where's real-world evidence falling short, or where would we like to see it do more?

And I think that there's a lot of vocalized desire for more real-world evidence use cases. But in practice, we don't always get there. So for example, if a regulatory body mandates a post-marketing study, follow patients on a drug for x number of years, sometimes regulators are receptive to modernized approaches that involve a more real-world, data-centered approach. Other times, we're reverting back to legacy approaches and more traditional clinical trial-like methodologies.

And so there's a little bit of tension there between, one, the stated goal-- yes, adopt these modernized technologies. Use real-world evidence to answer these questions. And then where the rubber meets the road in the level of rigor that either a regulator or a sponsor as well-- sponsors also fall into this trap of starting off with the idea of wanting to do real-world evidence research that get unadulterated, real-world data. And we find out later what's really desired is something that's more along the lines of a structured visit schedule where you're mandating certain visits. And that has its place as well, but that's where we see that tension.

ALEX MAIERSPERGER: Really helpful. And the sort of changing your mind midstream-- or I think it goes back to that fit for purpose, and that the question is so important to get the data that you want. You have to be able to know the question first. And so being able to design things in a way that says, what are we asking for, and what do we value most?

AARON BERGER: That's really-- if you really desire an unadulterated view of real-world experiences and treatment patterns and practices, or is there a desire to collect certain data at certain time points, and that cannot be sacrificed for the unadulterated, real-world view? And so really having a clear vision of that from the start before you design the study is absolutely critical.

ALEX MAIERSPERGER: You shared earlier the camera example of a photo versus a video. And it made me think of two statistics, one being, how much data is generated from pictures? And maybe I'll get this statistic wrong, but there's something like every two minutes, we take more pictures now than previous centuries.

There's also the fact that they say over 30% of all of the world's data comes from the health care and medical industry. And so obviously, the industry at large is generating massive amounts of data. You have electronic health records, claims, the various phases you talked about in clinical trials, modern wearables. What's the most important new signal that we can capture today within all of this data that we couldn't 10 years ago?

MASSOUD TOUSSI: Yeah. I would say there are-- I would like to talk about two different things, because there have been really two different signals which have been generated and that we can capture today. One, really a revolutionary event that I have been able to witness during the last 10 years, was the democratization of common data models. Because the common data models like OMOP now allow federated studies, and I was not seeing any federated studies in the past. And that has changed the entire picture because now, conducting a multi-country study is available for a team of two, three people.

Whereas in my past experience, past professional experience-- I come from IQVIA in the past. And I remember that IQVIA, with its multi-national databases, was the only one able to do multi-country studies.

And the other element that I see from my longevity perspective is all of these mature products, which are all connected, and which seamlessly collect information about our health, like these watches and also other tools, which are available and which capture continuous data and relay that to the other health information. I used to see from medical devices like pacemakers and all of these kind of other devices, which are permanently connected, I used to see those signals.

But from time to time when we were talking with-- for example, I remember a discussion with Medtronic. They didn't have other information to provide us. So this information was like one-dimensional information. We couldn't really do a lot of things with it.

But today, even a rudimentary connected watch provides you at least five, six different types of data points, which enable you to run algorithms and make very nice conclusions about the health of the person. And that information is also in a format which allows you to connect it to other data points, either through tokenization or common data models. And it gives you much more insight, which was not available 10 years ago.

AARON BERGER: And I may comment on this as well. So when I think about your question, new signals we can capture today that we couldn't 10 years ago, maybe even five years ago, it makes me start thinking about the EHR and how far we've come there, the Electronic Health Record.

10 years ago, if you wanted to get a patient's electronic health record for consented prospective research, the way to do that is you're going to sites-- and I'm talking about a US example here. You're setting the health care providers up as sites in a study. You're asking those health care providers to keystroke data for you from the patient's EHR into a structured EDC. They're doing that based on their analysis, the human analysis and structured fields and physician source notes. And that's how you're organizing the data to do, let's say, real-world evidence research.

Well, in today's world, we can set up-- and this is what we specialize in doing-- direct-to-patient study designs. We can go directly to patients, ask them to contribute their medical records, collect their medical records because of-- on top of the standards that now exist for data interoperability, FHIR standards and whatnot. And so we can obtain that EHR data.

Well, once we've obtained it, until very recently, it still requires a lot of human process to organize and structure that data. Getting it is one thing. It's great that we can now enroll patients, and we can do so in a scaled way through a direct-to-patient study, but you still have a massive amount of data that you need to deal with.

So what's something that we can do now that we couldn't a few years ago? Well, we can use natural language processing and AI on that massive amount of data to help us organize it into a study database and to glean outputs. So it's a massive leap forward in terms of being able to, to your question, identify signals that we couldn't before, or would have required a massive amount of human labor to do so.

ALEX MAIERSPERGER: You mentioned a US-specific example, and obviously operate globally. How hard is it to operate globally when the guidance differs by region? Working across all the large pharma players and sponsors, what's the most painful inconsistency?

AARON BERGER: Well, one inconsistency I'll comment on-- and I'm sure Massoud will have some comments on maybe regulatory frameworks. But we do have an inconsistency and a mismatch in the timelines in terms of Europe and US in terms of patient agency over their own medical record and being able to contribute it to research.

So here in the US, I can take agency over my own medical record from wherever I receive health care. I can authorize for that to be aggregated, and I can contribute it to an observational study, for example. In the EU, that's coming. It's on the way. It's part of the EU Health Data Space initiative. But it doesn't exist today. It's on the way.

MASSOUD TOUSSI: Yeah, I think there are a number of things, actually. On the data privacy, as Aaron mentioned, I think the United States is a single country with more states than in Europe, because in the United States, you have 50 states. In Europe, you have actually 27 states.

But at the same time, the overarching regulation, which is in the EU, in many situations is not actually binding. It is not overriding the local country law. That's the reason why you have a multitude of different regulations. And even when there is, for example, European law, each of the countries is allowed to define and should define their own local law, and sometimes it translates to something different.

At a global-- on the global stage, we have seen in the past five years more than 50 guidelines in real-world evidence that I would like to take both as good news or bad news. The good news is that, OK, regulators are paying attention, and they are actually moving toward providing guidelines. But the bad news is that, OK, that creates even further diversity.

Earlier in this podcast, I referred to the data quality framework. For example, FDA issued its own data quality framework. EMA issued its own data quality framework. Both of them talk about the same thing, but they don't use the same words, so this makes things more complicated.

So actually, the good news is that finally the IHC, or International Harmonisation Council, which initially in the 1970s helped harmonizing the regulation around clinical trials, has entered into this topic. And they issued the first guideline, which is the International Guideline M14, IHC M14 last year. And there are other guidelines coming. So that's the good news. And I think with this harmonization, it will become easier for everyone to generate evidence which is acceptable by everyone.

ALEX MAIERSPERGER: I just-- I feel like I got the 101 or 301 crash course to the industry of history and where regulators have been, where they're going. Massoud, you're writing a book, and you've described an evolution from data collection to methods, curation, and quality. And now you say we're in an era of convergence of statistics, epidemiology, and AI. And obviously, AI is so common that even before the questions about AI, we talked about AI. So what does this new convergence and this new era that we're in unlock?

MASSOUD TOUSSI: Yeah, I think that's an interesting question. In reality, most of the people who have a few years of experience in real-world evidence, they don't have a degree in real-world evidence, because real-world evidence is a new science, the first use of real-world evidence in the meaning that we are talking about, because before, there were a few elements, a few instances of use of this word in the publications.

But for example, there's a long-- there's a manuscript about lung disease, et cetera, which has used this term, but not in the way we are talking about real-world evidence. But the first actual real-world evidence instance dates back to 2006, 2007. So the science is quite new. It's very new. And people came from different horizons.

But in this very small period of time, roughly two decades, it has so much matured that right now, we have it-- I would like to say it is absorbing into its gravitational space other disciplines. So now you have-- in the RWE, you have people who are specialized in AI epidemiology. And for example, up until five years ago, even three years ago, if you wanted to learn about real-world evidence, there was no book about this. You had to just take a book about epidemiology, another one about engineering, or maybe other things. Or statistics, for example.

But right now, there is an emergence of a unique kind of unified discipline. And in this unified discipline, you have a mixture of statistics, epidemiology, data science, and more and more AI. And someone in future who's actually entering into this discipline or wants to find a job in this discipline, they will not be asked the same questions as the people who wanted to take a job in epidemiology, or independently in data science, or whatever. It will be something unique that I would like to highlight as real-world evidence and real-world data discipline.

ALEX MAIERSPERGER: It sounds the superhuman era of person plus AI is something that we need to train for, something that we're seeing converge, and just really exciting. Aaron, you talked about patient-directed trial design. You talked about patient matching. Where do you see AI making a difference now? Is it up front in the design? Is it during the process? Is it after? You talked about the post-secondary studies. And then where will AI not help? Where are you seeing a place where AI is not anywhere close yet?

AARON BERGER: Yeah. So first, I would start by going back to my example, for an example of where AI will help, things that it will allow us to do that we couldn't before. So if we're running a direct-to-patient study, we're going directly to patients. We're asking them to contribute their medical records. We're combining that with claims data. We're combining that with PRO or QoL data that the patient's contributing themselves. We're building this beautiful longitudinal data set.

So one thing I already mentioned is AI's going to help us organize that large amount of data so that we can perform analysis. And at first, it will do that in the same way that we run studies today, just by organizing that data into an EDC that allows us to then run our analysis.

But what AI will enable is a much greater level of nimbleness and flexibility in our ability to ask and answer new questions. So in today's paradigm, if we start collecting data, and we have a new hypothesis, and we have a new question we want to explore, one, we have to see, well, are we collecting that piece of data from the patient's EHR, for example?

Well, we're not collecting it. Oh, we have to update the CRF. We have to ask the data abstractors to go look for it and find it. We have to collect it and put it in. We have to update our SAP. We have to update all our programming and produce new tables for this new question that we want to ask and answer. All that takes a lot of time.

So I think what will happen is AI, because we have all the data, AI will allow us to become more nimble in doing certain types of ad hoc analysis and interrogating the data in different ways, and getting rapid answers to it. Which right now, it's a very rigid, structured process.

One area-- you asked, where will AI perhaps not help? Well, I think that-- I'll answer it this way instead. There will be areas where we are inhibited from using AI and from trying to do some of the types of things I've described here, because we will have a QA curve to climb here. Let me step back for a second.

In most cases, technology oftentimes reduces the cost or is a deflationary factor in the broader economy. When we see new technologies, it serves as something that at least decreases the rate of increasing costs of things.

In drug development, we've seen the cost to develop a drug exponentially increase over the years. Several billion dollars to bring a drug to market. At the same time, we've had tons of technology adoption. So technology adoption's not really-- you're correlated to containment of the cost curve.

And AI will be no different. In order to adopt AI, we're going to have to hire lots of QA people before anybody gets comfortable adopting AI. SAS is going to release features, for example, that allow us to automate programming for tables and figures, listings. SAS is going to spend countless hours validating that solution. And then everyone else that uses that solution is also going to spend countless hours doing validation as well. So we've got a big QA and validation curve to climb before we can start realizing some of the benefits of AI.

ALEX MAIERSPERGER: I think you acted as a great recruiter for the future of the industry of thinking-- just hearing the work that goes into, now you've got to update this table, and then you've got to go back and update this other table. And now that you have those two tables updated, you've got to go back and update this third table. And just the work that will become-- I love how you talked about just the creativity of AI, and maybe humans plus AI, of being able to say, what are the questions we aren't asking today? And what are the interactions that we don't have today that we potentially could have with this drug or with this design?

Just really exciting to feel as a patient that this is happening. And then to feel-- as this is someone who works in the industry, and this is going to be your future life and you're going to work on these breakthroughs, really exciting. Massoud, you've used an analogy that we trust autopilot on airplanes, but we hesitate with AI in health care broadly. What's the safe starting point for AI and clinical decision making?

MASSOUD TOUSSI: That's a very good question. There are actually two dimensions to this. One is using AI as the medical technology. So using for health care. And the other one is using AI in doing studies. And obviously, I would start with the second one, because it's safer. It's less harmful if you have any issues.

So I think in AI, we have two aspects. One is the technology, and the other one is the adoption and the economics around the technology. I would like to say why I'm asking-- why I'm talking about this.

Because interestingly, the first prescription AI tool, which was approved by FDA, was done by a company which got bankrupt just two years after they had the FDA approval. This means that we need to think about not only areas which are allowed, but also areas which generate enough cash flow to go to the next step, right? Strategically, we need to position ourselves in those areas.

So in AI, I would like to say a part of the AI, which is very mature, is the predictive analytics. So everything related to predictive analytics has already been tackled, and also natural language processing. This belongs to the 2010s. So already by 2020, you had all of those algorithms related to natural language analytics and predictive modeling, which have been already matured and kind of integrated already in the decision making of regulators.

The piece which stays and remains to work on is the large language models. And these large language models are moving with a very, very high speed. Even before-- even within three months, I see the models are improving in a very surprising way.

So I would say the only thing that the regulators can do, and somehow they have already tried doing, is to define frameworks, because they cannot actually catch up with the speed with which these models are actually improving. So instead, they define frameworks.

For example, the concept or the principle of human in the loop, which means that, OK, you want to do a study, for example, using AI, don't do it end to end. Do a piece. Have a human in the loop to check it out.

Let's say, for example, you want to build a system which builds the entire study from study concept sheet to the study report. So consider humans who would take your concept sheet, verify, and then give it to the next module, which takes the concept sheet, generates the protocol.

Then have a human who actually verifies that protocol, and then give it to another module, which takes the protocol and gives you the statistical analysis plan. And then for the codes, do the same thing. So that is a principle that I like very much, and I think it makes perfect sense. But I firmly believe that at some point, that human in the loop will also be another AI.

ALEX MAIERSPERGER: Ooh, that's a controversial opinion, maybe. What does that mean for the humans out of the loop, maybe? Does that change the nature of the role that the humans will play?

MASSOUD TOUSSI: Let me say something. I'm saying this. I have goosebumps on my arms. Let me tell you about a personal experience. Last year, I had a sabbatical, and I started working on an AI-- for example, an AI system which would allow building protocols, statistical analysis plans, et cetera.

And once I got finally the first version of the protocol, which was generated by the system, I couldn't sleep at all, because the main thing I was thinking about was, what are we all going to become? The AI is-- this AI generates something that I wasn't able to generate in terms of quality, in terms of extensiveness, et cetera, in terms of time efficiency, lack of errors, lack of editorial errors.

Whatever you could think about it, I think that I wouldn't be able to generate what this tool generates. So this is what I would like to say. I would really recommend people to read that Homo Deus book, because it is actually predicting what is going to happen later. Yeah.

AARON BERGER: Yeah. I mean, this conversation makes me think of-- and it's almost cliché at this point, because it's being said by a lot of people, but I think the theme of this comes to mind. AI won't replace humans. AI will replace humans that don't know how to use AI.

I think you can apply that thinking towards, let's say, statistical programmers, for example. And when we think about, what are the rate-limiting factors for conducting more research, for answering-- for asking more questions about the safety and effectiveness of a medicine and getting those answers? The regulating factor is the amount of cost, the amount of human labor that goes into producing that analysis. All the things that we've been talking about today-- abstracting data, updating programs to produce tables and figures-- all those things are a function of human time and cost. And that becomes the ceiling for how much research you can do.

Well, once people are enabled with AI tools that allow us to organize more data at scale and ask different questions and produce different analysis, because a single statistical programmer can now do two or three times x the work, well, now there's just-- that doesn't mean we necessarily need two or three times fewer statistical programmers. It means that we can do more research. So the universe expands. The pie gets bigger.

And so if you apply this thinking earlier in the drug development continuum, perhaps we can-- well, first of all, we'll be looking at more assets, more molecules through AI models. But we can bring more models into the clinic. We can bring more assets into the clinic.

We can do more phase II and phase III research. The phase II, phase III research we do will be more likely to be successful and produce candidates that progress through the pipeline. More treatments will make it to patients ultimately. So that's another kind of vision for how this could all go. It just expands the universe of research that can be conducted.

ALEX MAIERSPERGER: As a big fan of pie, I'm a big fan of hearing the pie expanding. And just-- it really is truly exciting to hear, apart from all the pure process elements of, we can ask more questions, we can design more trials and things, this ultimately means you can get better, safer, faster drugs to market. And we all are impacted, whether in our own lives or a loved one's lives, friends, community.

And so just knowing that some of the people that haven't been able to get the medicine that they've needed or things, that they'll be able to get that, and that there's going to be more options for more people is so incredibly inspiring to hear. I think, Massoud, I identify with the goosebumps feeling when thinking about the future of where health care and life sciences goes.

Real quick question of, where-- Aaron, you mentioned that-- you mentioned your colleagues in the real-world evidence world. And so for colleagues not in the real-world evidence world, what's a myth of real-world evidence that you wish the industry would retire?

AARON BERGER: A myth that the industry should retire. Well, I guess for-- oh, for those not steeped in real-world evidence. Well, most of the time when I am talking to people about what we do at UBC, everyone's mind immediately goes to controlled clinical trials. Oh, you study medicines, you run trials.

And yes, we do run trials, but we also run the types of programs I'm talking about. Direct-to-patient studies, following patients in real-world settings after drugs are approved. Sometimes it's tied to a drug that a sponsor has hired us to run a study on. Sometimes it's simply a disease registry.

And so people not in the space are usually a bit surprised to find out about the amount of research that's actually being done on drugs that are out there in the market, have been in the market for years, and the amount of research that's being done on diseases that's sponsored by industry, by pharma, to learn more about treatment patterns and how different combinations of medications and how disease progression is occurring out in the real world. So that's something that's kind of eye-opening for people that I talk to that aren't familiar with the space.

ALEX MAIERSPERGER: And then for either of you, what's one thing listeners should watch for in the next 12 to 18 months-- so in the near-term future that signals real-world evidence is making our health care system and the broader ecosystem a real true learning system, not just an add-on to trials or part of the design?

AARON BERGER: You know, I'm not sure there's going to be one thing, to be honest. It's a-- we've been on a long journey, and we're going to continue to climb that adoption curve. I think we're going to continue to see real-world evidence applied in more and more use cases. It's really-- it's taking hold, and it will take hold even more for safety purposes, for long-term safety studies and long-term post-marketing efficacy explorations.

We'll see it take hold more and more in indication and label expansion and reimbursement decisions from payer bodies. So I think we're just going to see that we're going to continue to climb that adoption curve.

ALEX MAIERSPERGER: I love the inevitability of, it's taking hold, and will take hold. And it certainly paints just a bright future for health care and for life sciences, and for the drugs that are coming to market and that will come to market, the ones that we know about and the ones that we don't know about and will know about here in the future. So Aaron and Massoud, thank you so much for joining The Health Pulse Podcast.

MASSOUD TOUSSI: Thank you.

AARON BERGER: Thank you.

ALEX MAIERSPERGER: Thank you for listening. To prove you're still human and in the loop, please email us, [e-mail]. We'll see you next time.