Transcript
Speaker: Welcome to Clinical Data Talks, a podcast brought to you by CRS-Cube. I'm your host, Sylvain Berthelot. Join me and industry experts, how we discuss the latest trends impacting the world of clinical data.
Speaker: Today, I'm really pleased to welcome Dr. Kavita Gade to the podcast. As a clinical data management lead, Kavita works at the intersection of clinical data, technology and strategy.
Speaker: She brings a thoughtful and practical perspective on how data management is evolving in our industry. In this episode, we're going to explore the real challenges data managers are facing today, and we'll discuss if technology is genuinely solving solving problems for data teams.
Speaker: Kavita, welcome. I'm really glad to have you with us. How are you doing? Thank you, Silva, for the kind introduction. I'm i'm just fine. it's um It's sunny in Berlin. It can't get better. Thank you.
Speaker: Nice, nice. Sun always makes things better. So before we we talk about the challenges, where we're going to to focus on challenges, some of the challenges for data managers. We're also going to talk about what you see in technology. ah But I'm curious about your current setting and what your day-to-day work looks like. So could you share with us what a typical data flow involves for you in precision medicine?
Speaker: Right. so if fee i would I would like to not just start focusing on precision medicine right now, but provide a broader perspective to data flow to start with. So currently when ah Now we are dealing with the digitalized data and the due to this digitalization, we are talking about data flow, which is now non-linear. We are dealing with this data mesh coming from all all different sources.
Speaker: And when I say sources, it is clinical EDC system, then clinical labs come into picture. We have device data.
Speaker: We have patient reported outcomes. And in precision medicine, when we move more deeper, then we may talk about imaging data, this big DICOM analysis, then this multi-omics data, bringing the complexity of not having a ah linear data pipeline flow,
Speaker: And in a typical day-to-day life at the moment in current environment for a data manager, you are dealing with this different vendor, which are having different data environment and this heterogeneous data coming at one point in in one place and really talking in different languages.
Speaker: And this forcing, like as a data manager, you have to at the end force all these different source to talk in one language, which will be the C disk standardization.
Speaker: And really making these two end meet is becomes the day to day life of a data manager in clinical data flow environment.
Speaker: Yeah, so the the the big challenge I imagine here is not only aggregating the data, but at the same time, when you say talking the same language, for me, what came into my head straight away is time points. How do you know if the like when to look at those data points and if ah ah another data point from a different source has a similar time point or not. Is that a challenge as well?
Speaker: that that that is ah That is a challenge, but I think um we we have started talking about this a bit um bit early, but um the early the better, let's say it.
Speaker: um this is This is something where this semantic layer basically, introducing the semantic layer comes into picture. And then we are talking about the...
Speaker: the data latency where data is there, for example, but it is not available for analysis due to the fact that um what we think as a time point, for example, may not be the same time point in the vendor data source. And that requires reconciliation process. where we have to align, look, these time point do not match and we need to fix that. and This two-way exchange of data makes the data latency even more longer. Is that a challenge that is improving or is it getting worse over time? You talked about the fact that your data flow is not linear anymore. So is is it that it's become more complex? complex
Speaker: Yes, it has become more complex and especially the complexity, at least in in my experience, what we have seen is that post-pandemic, we had this a very unique thing which got introduced, um decentralized trial, for example.
Speaker: And having those decentralized trials that brought in more digitalization, introduce introduction of more variables there.
Speaker: And so I wouldn't say that the challenges are reducing, but what is happening is that the infrastructure it has remained static, relatively static in comparison to the digitalization, what we are bringing in. So we are dealing with 21st century data and digitalized technology, but that integration of that 21st century
Speaker: data is getting integrated into 20th century infrastructure still. And and that's, i I think that's that's very, um it that that I wouldn't say that baffles me, but that is something which ah which you see everywhere. So um you will see at one point that one vendor is offering those services which are state of the art service. But still, you have to really consider that will your infrastructure be capable of using those technology. So this understanding of um but that, that the system is capable to handle it.
Speaker: and you know, and and then using that vendor, using the technology that will make things a bit easier. And but I wouldn't say that things are getting easier and simpler, but we just need to make wiser decisions there, which things and which vendors and which technology we have to use based on the infrastructure we have in place.
Speaker: It's very interesting and i like how you put it, that we have brand new technology that's very advanced and and but but you're trying to build this on technology or infrastructure that's not as advanced. And yeah, it's a very interesting perspective.
Speaker: And I imagine that so it's a very tough challenge to have, actually. ah But that that reminds me of something that really um pains me in a way in the industry is that...
Speaker: um I feel like in some ways we're always looking at the shiny object. We always want the latest innovation, but that we're not necessarily addressing basic needs. And I think that relates to what you're saying, that the basic need here is having the infrastructure evolve at the same time as technology, but we're not doing that.
Speaker: um So you've hinted at integration layer, you've hinted at infrastructure. So for you, how would you like the the data flow to evolve and what tools would you need for that?
Speaker: Right. i I would really, um so yes things like this, you can't have, there is not one solution which you just get off the shelf, right? Saying that this is um ah like like you go shopping and then, so it's for in clinical data because it's so it's it's so vast and and we are, so basically the aim should be having the mindset of data being an asset and not the byproduct of the clinical trial, plus that having and really using that asset as as ah analysis ready data set.
Speaker: So now when we say that there won't be one off the shelf solution available and you have to have a a solution which will be in mind, which will be scalable at time depending on the needs and then when selecting the vendor right from the beginning. and Again, it's not that um when when we say, okay, we we know that we have selected this and select selected the vendor and this is this vendor is going to be the best and will last in through our whole trial. But this having this scalability in mind and then flexibility in mind is is very important that you cannot just have one fixed solution of a problem. At times you will have to be flexible and so you you need to select a vendor which will provide this flexibility depending upon because what happens in clinical trial in precision medicine, let's let's talk about and precision medicine or in oncology. that you you start with something but depending upon the adverse event and and the severity of of the um of the which which we identify in the trial you have to amend those end points and those end points will then affect the um
Speaker: the way we are selecting or recording our results or endpoint and that flexibility should be kept in mind. So there is no one solution which we so we should say that, okay, this is what we have selected, we will go with it and that will become the um that will become our end solution for that.
Speaker: So that is what I have learned through this throughout my my experience that this flexibility offer should always be there on the table.
Speaker: I imagine how difficult it is because you don't know what's going to happen. So so you could, and and I hear what you're saying, you need to choose your vendor based on including as a selection point, the fact that they need to be flexible, but you don't know what this flexibility needs to look like in a way, because you don't know what each trial is going to throw at you. So I can imagine that that it's a big challenge.
Speaker: ah So going back to to to this infrastructure, ah so So you talked about the fact that it's not in line with the expectations of of the data you're getting.
Speaker: um you also mentioned data ah not being a byproduct, but essentially what you want as a data manager, I think that's what you said, is it having data that you can analyze straight away.
Speaker: But that's a big change compared to what you are working with at the moment, like all those systems that don't necessarily talk the same language.
Speaker: How do you think you can achieve this big change? Is it possible and and at the same time running clinical trials? No, I, I, it it won't take, we won't be able to implement this from one day to other, Silva, to be very honest, because had this been that easy, everybody would have implemented it so In our industry, we do see that um bigger bigger players or bigger pharma companies coming up with some solutions, for example. But we have not seen yet that those companies, they have that solution implemented for
Speaker: like like or through their clinical trial um supply like clinical trial data chain let's let's put it that way so they might use um some domain specific thing for example some some might take just decentralized trial and then try to implement it and then some might do it on one therapeutic area so these pilots what they are running they are run always on on a smaller scale. And when you bring it to the real scale, you see that system start crashing at that time. And then you will have to really bring in that um
Speaker: um fire, and like fire extinguisher and then to extinguish, okay, this integration doesn't work. We didn't get that prompt that prom The API doesn't work, you know, for for this this big influx of of that the data, etc. So at one point, bringing that pilot to a bigger scale will always be challenging.
Speaker: And and it will take it will take at least like one or two big, big trials where people will or a data manager will volunteer to let them implement that one pilot on their study because what they will be dealing with is um running the trial smoothly and dealing with those small fires which is occurring every day so i i to the clear answer is i don't see it happening from one day to other but it is a single process and then we need to have
Speaker: there a mindset change, which is which is very important. So if if a mindset doesn't change, because what happens in clinical trial is that clinical trial is not just the responsibility of a data manager. Clinical trial data is not just a responsibility of clinical data manager. But there are so many stakeholders which come into picture, like we have safety database, we have pharmacovigilance data. They would like to select their vendor according to their needs and convenience. Then we have this dosimetry, DICON people coming in. They would like to select that vendor accordingly. And it goes on. and one and And if this mindset is not there, that at the end we would like to have this unified system which will which should be able to talk with and all these um variables should be able to talk with each other.
Speaker: um it will It will take even longer to have that one solution in place. And again, that will require this mindset change and keeping this in mind that data in clinical trial is an asset as a whole, not just that domain is an asset for for that specific function. And that brings in the biggest challenge because what I have seen is that um clinical data managers they are involved in vendor selection for their EDC system, for example. Safety selects their own vendor.
Speaker: Then other function will select their own vendor and and this goes on. And then you are again, ah every trial you are talking about the same problem and it goes on and on.
Speaker: Yeah, and that's a very good point you're making here. The fact that you there's not one single person in charge of selecting the whole technology. and And I think that's what you refer to in terms of mindset change, that ideally you need a holistic view of your whole technology stack so that it makes sense and it's easier to manage.
Speaker: um Something I always wonder, and you've mentioned large pharma in a way trying to solve that problem, is that large pharma can potentially ah spend a lot of time, throw a lot of money a problem like this.
Speaker: But I wonder if it's something that can be replicated in smaller sponsors. Do you know if that's possible? Or do you expect large pharma to solve a problem and then it becomes easier to implement for smaller sponsors?
Speaker: Right. So these are these are the things where financial financial please our finances plays a very big role. when we are when When we talk about bigger pharma, they have their own IT team, which is capable of implementing or providing the solution. They will be able to take or hire those cloud solution in those partners who can provide or deal with this large language models, etc.
Speaker: But when it comes to smaller pharma, um the the infrastructure and like like the basic IT infrastructure which is provided is is is not at a scale for to to making it that scalable basically. and And they always try to then go for an outsource solution where ah where they would contact a so CRO which will basically provide this whole infrastructure, et cetera. Because smaller pharma, they would their aim will be totally different. At their at their um my mindset, it will be like, okay, we need that one blockbuster. And then let's just have that one blockbuster ready to get that um cash flow in and then implement our solution, right?
Speaker: um If those finances are not in place, it will be, we we cannot really compare those big pharma with with smaller biotech ones. And there the need is really to take these smaller steps that selecting the right EDC vendor, for example, to start with, because many pharma, many of the smaller biotech, they might not even have in-house um edc providers basically so they will be solely relying from end to end on this on the cro to provide this complete solution there because that that makes the everything easy in a sense that they do not they don't have to deal with it but the so cro will be dealing with it but at the end those smaller biotech will be dealing with that ah cro and ultimately dealing with that infrastructure not as their own but as some but something which is hired like a rented car let's say it
Speaker: Well, yeah, and and I imagine that then if you have multiple trials, not necessarily with all the same CROs, that's when the need comes to do your own thing, because i imagine that it becomes quite difficult.
Speaker: ah Do you see ai as a solution to solve the challenges we've been talking about?
Speaker: i It will be very unfair if we would not have talked about AI because that would not have made our our talk very fancy. So everybody talks about ai Selva.
Speaker: And ah it's not it's not that new, but the implementation, what everybody is trying to do with ai is is relatively new So it started with machine learning, with automation, etc. And now there is this product. For example, we have this agent, AI and and whatnot.
Speaker: So AI as a solution to our clinical trial, look, we are not able to solve the infrastructure related issue there. And AI needs infrastructure where we see this big player like NVIDIA and all these coming in, providing this infrastructure to make this large language model run smoothly.
Speaker: So To be for for AI to work properly, we need proper governance, proper infrastructure. And and the biggest thing is a unified, like ah a very unified layer of data.
Speaker: Because if you won't if you will have a heterogeneous data, AI will not be able to function on that heterogeneous layer or if even we we somehow manage to make it work on that heterogeneous level level, it will not work for impact. It will just work because it has to work, but it won't give any any results or or desirable results, let's put it that way. So AI cannot be worked ah cannot be used as a magic wand there. I don't i don't see it.
Speaker: um being being used, but then specifically on a very specific pilot, let's let's put it that way, again, running a pilot on on a small decentralized trial or or a small rare disease model, for example, where we have one unified system, one unified um data, proper stratified data, and then see if AI works there.
Speaker: And AI will then be used for impact to provide us this insights on data, which is getting and and where this inflow is coming, where we can avoid this redundancy of work reconciliation, multiple reconciliation cycle, and then really prompting for um data points which which will look very unusual at the end when the analysis will be performed, but those AI can be used already there. But just say, okay, we ah we have whatever data platform, let's just integrate it with AI.
Speaker: it will It will make life more harder than making it easier, in my opinion. Yeah, I get that because yeah it goes back to the infrastructure. If you don't solve the infrastructure, no matter what you could put on top of it, it's not necessarily going to solve your problems.
Speaker: um But it's interesting. um it The more we talk, the more challenging it sounds. and But I'm an optimistic, so I'm hopeful that we'll get there eventually.
Speaker: and I just have one last question for you. What's the best piece of advice you've received that you consistently apply at work?
Speaker: i I had very nice mentors when I started my clinical trial, Silva. and the biggest and and the greatest advice what i have received so far is having that open um so open mindset because again um the whole talk started with we dealing with this amazing digitalization and deal and trying to integrate it with um obsolete, let's say a bit of obsolete infrastructure. So if we are not coming up with an open mindset and then this flexibility, how to really
Speaker: think of ah and this problem solving ability that is going to make everybody's life difficult, especially in clinical trial and and and keep on learning these new technology and identify those loopholes there and see, okay, this is some something and really keeping an eye on those loopholes. See, okay, this is what is going to create problem in a long run.
Speaker: and and keep keeping an eye on those things. So ah really a learner's attitude, open mindset and and and and just jump in those open ocean of of clinical trial data flow and that's the only way to to deal with um our hit our really complex data environment and it's getting complex and like the more the moment we are talking right it's not getting simpler it's getting more and more complex every day
Speaker: Yeah, I like that. And I agree with you, especially in our industry. If you don't have an open mindset, I'm pretty sure you you'll find working in the industry very challenging. But that's what excites me that the every day there's something new and and something interesting. And you always have to learn. You can't just sit there without evolving with it.
Speaker: I loved our talk. Thank you so much, Kavita. We've talked about challenges, but you've given me ah more reasons to to keep doing what we do in the industry and and challenge the status quo. So thank you for that. I really appreciate it.
Speaker: Thank you. Thank you, Silva. It was a pleasure talking to you. And again, it's... it's some i Every time i i i come to work or start working, it's just,
Speaker: i it's it's it's it it never gets boring. Let's put it that way. So I'm i'm glad that I work in this um in this human-centric environment. And I hope that there will be many more years where I will be able to contribute in this environment.
Speaker: um to to this noble cause of clinical trial and bringing impact to human life. Thank you. Thank you for giving this opportunity. That's a perfect message to finish on. Thank you.
Speaker: And thank you all for watching us. You can find more clinical data talks on the CRS-Cube website.
