IQVIA Holdings Inc. (IQV) Earnings Call Transcript & Summary

Hello. I'm Jeanne Northrop, Managing Editor with Applied Clinical Trials, and I'd like to welcome you to today's live broadcast, how to achieve comprehensive oversight of your entire portfolio of clinical trials, sponsored today by IQVIA. If you are a small or mid-sized pharma company without that simple solution to continuously consolidate multi-force CRO data into standardized views for analytics, you may be left in a bind. You need current data from vendors and actionable oversight, but often only get these insights via vendors' monthly reports. The variance in data cadence from multiple vendors means you are typically working with some data that is relatively new and some data that is older by a month or more, which we all know is a hardly comprehensive picture of a trial to reassure your executives' oversight concerns. IQVIA's clinical data analytics suite can solve this challenge and give customers easy-to-use dashboards and a real-time view of critical KPIs regardless of who is conducting individual trials. In today's webcast, you'll learn more about this self-service and vendor-agnostic platform. We have 2 great presenters on, and we look forward to kicking off this presentation. However, just a few important announcements to make before we begin. This webcast is designed to be interactive. And so we encourage you to ask questions to our presenters throughout the event. You can submit your questions to them by typing them into Q&A box. That can be found at the bottom of your video player. You can enlarge the slide window by clicking on the small icon in the bottom right corner of your media player. Please note, however, that all slides will advance automatically throughout the event. If you do happen to experience any technical problems viewing or hearing this presentation, click on that question mark located in the top right of your presentation window. Now I'd like to welcome Wendy Morahan and Gary Shorter both of IQVIA. Wendy joins us with 25-plus years experience in the life sciences industry with the career spanning academic research, preclinical drug discovery as well as clinical trials culminating in a focus and passion for delivering technology solutions that can help bring treatment to patients faster. Wendy is currently part of the product strategy leadership team for IQVIA's clinical data analytics suite, providing software-as-a-service solutions for the market as well as IQVIA's internal CRO needs. Wendy is responsible for strategy, product management leadership and go-to-market activities. Gary pursues the use of emerging technology to provide new and more efficient capabilities to enhance clinical trial management. This includes the development of new design software that supports recent advances with AI capabilities. Gary's team has developed several micro products and micro services that can be plugged in and used by any software-as-a-service solution. So Wendy and Gary, welcome. Thank you so much for joining me today. I appreciate your time. We're really looking forward to this presentation. Wendy, feel free to begin the presentation.

Jeanne, thank you so much for the nice introduction. I am going to quickly share my screen to get us kicked off. So again, thank you very much for joining us today. I think this could be a very interesting conversation. Please feel free to ask us any questions in the question chat pod. Happy to have discussions around any of these topics around how to really achieve comprehensive oversight of your clinical trials. I think, first and foremost, I wanted to set the stage about IQVIA Technologies. I think probably everybody is familiar with IQVIA as the world's largest CRO organization. But indeed, we also have a business unit called IQVIA Technologies. That is where Gary and I sit. That is where we do our work. And IQVIA Technologies is really focused on those technology offerings, both to the external market as well as for supporting our internal CRO business. We have technologies in clinical, real-world, compliance and commercial across the board. Now our focus is absolutely clinical technologies for this conversation. In fact, we'll be talking mostly about data and analytics and sponsor oversight of clinical trials. But IQVIA Technologies also has a very broad swath of clinical technologies that deliver insights, help with study planning, site management, kind of everything you can think about soup to nuts on a clinical trial. We're also supported by 13,000-plus technology experts. So when you think about IQVIA as the world's largest CRO, even taking a small slice of that for IQVIA Technologies, still results in 13,000 technology experts that are working on these different solutions and working across 300-plus sponsors. So a lot of experience and insights that we are lucky to be able to gain from working with the smallest biotechs to the largest pharma. So for today specifically, the agenda, and you'll see that I've given us -- split it into 2. There's 2 speakers, 2 parts of the agenda, makes complete sense. The first is, I will be covering challenges in clinical trial oversight. So first of all, what is clinical trial oversight? What's required? Like what the regulation say is required. What are some of the challenges that we face with clinical data today? And then talk about really the value of having the right clinical trial data oversight solution. Then I'm going to hand it off to my colleague, Gary, who is going to focus on the impact of AI on clinical trial oversight. What does it look like to automate clinical data flow and automate the standardization of data with all of the new focus on generative AI technologies like ChatGPT. How can generative AI innovations start impacting and adding value to oversight? And then what's the value of intelligent dashboards? So descriptive business analytics and business intelligence is extremely important for oversight, but adding intelligence to those dashboards is the next best thing, and Gary is going to review some of that with us. So key takeaways. We're hoping by the time that we're done today and we're done answering questions, you've kind of started to understand or maybe even just feel justified in your knowledge about the challenges you're experiencing, like, what are all these challenges that we have with oversight with all of our data? What are some of the optimal ways that we can improve trial oversight? And how can a platform for clinical data and analytics and AI automation benefit sponsors? So let's take just a quick minute to talk about challenges and some of the regulatory requirements that are really driving our focus on sponsor oversight. Starting with challenges. This is a slide that I use often in the data world and in data and analytics for the last several years. I can start almost every conversation by talking about we know that the world of clinical trials, the world in general, not even clinical trials, the world of data in general is just so complex and ever expanding. The expansion and innovation and technology that produces new data sources day after day is just overwhelming. And that's true in clinical trials as well as in the broader real world. This leads to this massive increase in data; in the complexity of the data, in the volume that we see it in, the velocity that it's coming at us. And so it's just this massive shift in the avalanche of data in our lives. Data is also everywhere. I don't know how else to describe the fact that it's not just about the core concept of a decentralized trial. That is a thing. But there's also this decentralization of information even outside of a decentralized trial in any clinical trial. It's no longer a world where everything is focused on EDC and maybe labs. It's a world where data is in IRT, eConsent, eCOA and connected devices and ADC and labs, but it's just all over the place -- and electronic health records, I didn't mention EHR. But that makes it very hard to manage and it makes it very hard to oversee that data. And so we really have to start taking a look at what can we do to bring that data together, to make it easier, to get a consolidated picture of data for both -- for all sorts of use cases, whether it's the tactical data management when you're conducting a clinical trial or whether it's oversight, like I just need to get my data consolidated and then I need to get my data from various CROs so that I can have oversight of all of my trials. We also see that it's very difficult for sponsors to kind of maximize the value of existing data. There are so many tools out there today with the AI models that Gary is going to talk about. And oftentimes, those models come with kind of requirement of unique training data. You need a pretty strong data foundation to build those models and train those models based on your sponsor experience and your sponsor trials. And oftentimes, that data is at hand, but it's unstandardized. It's in various formats. It's in various locations. And so being able to consolidate historical data and then utilize that data to inform future trial strategies and future trial conduct is a huge challenge, but I think one that will bring a lot of value to sponsors. And then lastly, see large challenges in just the slow adoption of AIML solutions by sponsors. It's -- we're in a quite conservative industry and it is definitely something that we're not just able to snap our fingers and start using ChatGPT in some of our tools like we can in our real maybe everyday lives. So we have to really look at how can we help sponsors accelerate their adoption of some of these AI technologies. We'll focus in just on a little bit pieces of these challenges today, but this is kind of how I see the whole world of challenges when I think about clinical data and analytics. And then specific to the oversight conversation is our beloved ICH E6(R3). And I wanted to focus in just a little bit on one section of that. There's an entire section of ICH E6 called Sponsor Oversight. And nearly every section -- I was looking at it this morning, nearly every section starts with the words, the sponsor should ensure. Some of the sections don't use those exact words, but most of them do. Things like the sponsor should ensure end-to-end trial design, conduct processes are undertaken with sufficient quality for trial results, patient safety and decision-making. The sponsor should ensure compliance with protocol and regulatory requirements. The sponsor should ensure deviation classification. Sponsor should ensure risk management. And I think this was one of the -- more of a guidance than a sponsor must ensure, but there needs to be this idea and guidance around you should look at what committees do need in place for oversight. And so I think that the reason I wanted to show this and why I kept saying the sponsor should ensure is when we start talking about clinical trial oversight, I think some of the most impacted sponsors are those sponsors that don't have the resources or have made the decision to work in an outsourced CRO model. So I would consider those biotech, early phase, even up to mid-sized pharma are oftentimes working in a world of a highly outsourced model. And that can be the model where it's really hard to actually have the sponsor oversight that you're required to have. So yes, the CRO has oversight of the trial, but at the end of the day, it's the sponsor's responsibility and the regulations make that clear. So now that we know all the scares of the challenges and the requirements and the regulations, it's like, okay, so how can we start looking at technology even as sponsors that are outsourcing our trials? How can we look at technology that we can own in order to start ensuring our ability to have oversight of our trials? And that's what we're here to talk about today with clinical data analytics solutions from IQVIA Technologies. So CDAS, Clinical Data Analytics Solutions, you'll hear we start calling it CDAS for short. This is really a whole platform of data and analytics technologies that are focused on harnessing research data and transforming workflows. It's very modular, very flexible. So it doesn't mean that this is a giant platform that sponsors have to onboard. This is a very flexible set of tools that allow you to harness data, whether that's data from trials you're running or data from trials that sit with CROs, generate insights from those data and drive intelligent action. At IQVIA, we have a concept called Connected Intelligence. And it's the right data, the right insights, driving the right actions. And so you'll see that data insights actions focus here with the clinical data analytics solution applications. So what do I mean by this platform? Just as a quick overview, CDAS includes clinical data ingestion, acquiring data from anywhere, storing that in a data warehouse and then having tools on top of that data warehouse that allow you to standardize the data, map it into models that are supportive of things like SDTM data submissions or things like a reporting layer for standard, consistent analytics across your trials regardless of who is conducting that trial. So really the message here is tools to standardize disparate data so that you as a sponsor have the same type of view on that data regardless of where it originated. So once we've got that data in hand and we've got it in our data lake and standardized, now we can really start surfacing the value. We have tools around clinical data review which is truly that data management perspective on reviewing data and managing issues on a consolidated data set. Clinical analytics is where we're going to zoom in today. This is really about intelligent insights, driving intelligent action and appropriate level of oversight. And then lastly, something we won't touch on today, but our data science workbench is about exactly what it says, data science, that ability to make new connections between different data to explore hypotheses and potentially develop new unforeseen insights. All of this is meant to derive and surface insights that allow you to then do things, like for IQVIA, we develop what we call intelligent applications from these insights. But giving you this whole picture, the ingestion and mapping pipeline, those data surfacing tools in the middle, all the intelligent applications to the right, this whole picture leads to several use cases. And those use cases range from things that solve all the different data challenges I was talking about previously, but to really make sure we have the focus that we need. What we're here today to talk about is one simple use case for this type of data platform in the hands of sponsors, and that is consolidated data for operational oversight regardless of where your data is. So this is kind of answering the question, hey, I want real-time access to all of my data from all of my studies, whether they're insourced or outsourced. And I want to be able to produce consistent dashboards for executive reporting or for day-to-day oversight and management with my CRO. When we look at this and I think about kind of the numbers associated and the benefits of having a tool like this, we see that kind of 5 to 50. It's a really broad range, I understand that. But it's not 1 and it's not 100. There's this sweet spot of the number of active trials that we see our small to mid-sized sponsors engaged in every single year. And of those sponsors, you see about 76% sponsors are outsourcing some or all of the components of the clinical trial. What that can result in is weekly or monthly receipt of operational information about your study. And I've made that really broad as well because it's like the story here is one CRO may be sending you weekly feeds, another CRO might be biweekly or monthly, one CRO might be Excel spreadsheets or a PowerPoint or a PDF. The idea is you're getting data not often as quickly as you would like and in varying disparate formats that make it hard for you and very manually intensive for you to pull that data together into internal reporting and an oversight analytics that allow you to have more efficient oversight across all of your trials. And so the minutes to days number is, if you think about having a data pipeline that allows you to consolidate data across your various CROs into a data warehouse that already has the data model that you need for reporting, that means you can ask your CROs for more frequent data feeds into your system and then out to your clinical analytics in minutes to days, depending on your transfer frequency with the CRO. So it can greatly enhance your ability to see data quickly more often, more real time, more temporaneous data. So how can I see my -- oversee my portfolio of active trials regardless of insourcing or outsourcing model? This is really by aggregating, standardizing and storing all of the relevant operational and clinical data across your trial portfolio in the same way, and that's what CDAS clinical analytics specifically does. Now I mentioned earlier, the clinical data analytics solutions applications are very flexible and modular. Now I'm zooming in on, you have the data pipeline in place, now you just need clinical analytics on top of that. You don't need data management, you don't need the data science work venture intelligent applications. You just need clinical analytics to focus on oversight. Our clinical analytics tool is a business intelligence tool that has role-based dashboards, a whole library of pre-configured visualizations, and I'll give you a little bit of insight into in the next slide. But it also has self-service reporting, which we found is really critical for our sponsors. Our sponsors want to be able to not just build their own dashboards, but even take our standard dashboards and potentially brand them to your look and feel or remove a particular graph from a dashboard that's not relevant for you. So a lot of configurability and self-service ad hoc reporting in our tool. And of course, this is meant to deliver trial management, intelligence, monitor patient safety, generate intelligent insights. This is just a quick view. Obviously, I can't go through all of this in our short time together today, but what you would see in clinical analytics, you've got the my dashboard, which is self-service reporting. You have clinical dashboards, which may or may not be relevant to sponsor oversight. For sponsor oversight, you may not need all the clinical data, but there's going to be some things in clinical data that you are going to want from EDC like adverse events and deviations and medical history or CONMEDs. That type of thing that you're going to want to see from your clinical perspective in order to have the appropriate oversight. We've also got the study overview, which I would say, is more of an oversight dashboard across your portfolio. Total number of studies, studies by status, therapeutic area. Even dashboards that will show you studies by CRO, like where do I have my various studies with KPIs, milestones, study start-up and conduct dashboards, a dashboard specifically focused on protocol deviations and then of course site health. So a really robust kind of a library of visualizations and analytics to start with, and that's before we even start talking about some of the intelligence that we can layer on top of this. So if I circle us back and I think about the clinical data analytics solutions and that connected intelligence story that I told you, the right data, the right insights, the right actions. In this context of sponsor oversight, it's really about the right data, being all your operational data, CTMS, IRT, other sources, the ability to get subject recruitment data, both planned and actual milestone plan and actual data, site activation status, monitoring status. How are the monitoring visits going. And clinical data, like I just mentioned, PDs, AEs that type of thing. Having all of that data are allowing you to generate insights like you can even look at portfolio and budget status, by the way, but identifying insights like site activation is blocked. The critical one that I'm sure we're all aware of and have experienced in our lifetimes in clinical trials, the dreaded enrollment is delayed, which seems like that's always the case. But the sooner you can identify that and see that insight, the sooner you can be working with your CRO to correct and take the right actions. You can see milestones that are at risk. You can see site performance indicators and CRO performance indicators. So to have a CRO is actually at risk with some of the conduct aspects. Again, leading to the right actions, what do I need to do next, retrain the site, get additional sites that maybe we're on a selected list, but not activated. Start activating new sites, what CRO governance actions might I need to start taking. So again, getting you the right data, driving insights and clinical analytics and allowing you to take the right actions quicker than you could in a scenario, in a landscape where you're waiting days, weeks, months potentially for the data to get into your hands. So with that said, it's my pleasure to hand off to Gary Shorter and start the conversation about some of the cool AI and intelligent dashboard stuff. So I'll stop sharing and hand it over to you, Gary.

Yes. Thank you, Wendy. Hopefully, everybody can see the slides. So yes, there's a lot there from what Wendy has already indicated around the oversight and the clinical data. Let's get into where AI really sort of helps us support the oversight. So what we see here in IQVIA is really around these key areas. AI isn't just throw an algorithm at something and see what comes out. There's a lot of things where you can use AI and other tools to help you. So for example, in automation, the first thing is really making the flow of those clinical insight -- oversight information coming through to consume it and to help consume it easily by being able to recognize it, being able to standardize around it, as Wendy was saying, with some mapping to be able to code based upon what it's seen historically and being able to bring those ontology sets into it to be able to do with unstructured text and to be able to make it into more structured information. So there's really a lot around the managing of the data is coming through and being able to automate a lot of that process. So it's getting in without having to consume a lot of people's time and resources in that space. So that really is about that. But in addition, there are other factors around automation that help clinical trials, such as the operational activities coming through and efforts that you have to collect regulatory-wise such as all the documentation. So a lot of how can I automate being able to recognize what's coming through to me as a clinical trial user, as an operator and to be able to see where there are issues on the clinical trial flow or indeed on content flow and other flows of operations. If I can establish that upfront, that's helping me to now get into the next tools, which is around obviously detect the key information. So what are the key informations around the subjects, the site risk, those kind of issues, but also in being able to key, identify other pieces of information that can be important such as am I collecting the right information as content? Are there any issues with the site of what it's sending through to me? That can be an important factor of quality and compliance and regulatory needs. So it is about being able to have that next buffer of tools that help you really sort of feed that in. The third buffer obviously is then to say, okay, there are also -- being able to say, well, what's beyond detection of I found outliers, great. But can I also with my historical content and my knowledge of historical trials, be able to start to predict. So preemptively planned for the effort of, okay, there's something here that could become an insight or an important aspect of the clinical -- running the conical trial. For example, things such as being able to look at predicting the future of the site information and any protocol deviations, being able to look at early signals for safety issues or risks. Those are important factors that we believe are worthy of having the algorithms and tools in there as a library of functions that help feed that oversight, but also to look at and recognize false positives and be able to correctly manage those in a productive way. And then finally, as we all know, and I think there's nobody that can avoid this, the ChatGPT and the Generative AI. We do see them as of value. The question is where do you put that? And so it really is about being able to utilize those tools in a way that suits best to the clinical trial, whilst maintaining the regulatory rigor and compliance that's needed because we know that these tools are -- have been talk from a very generalized form and what you need to do is make them more domain-specific into those areas to really be able to feed that oversight correctly and properly and to not give you issues with the use of the AI. So just moving on. The automation use cases, just to show in a few examples around that. So as I mentioned, there's the mapping. There's also other areas where we can do recommendations around queries, because again, you've got the historical information to be able to identify that. There are recommendations around. Because I've seen this structure setup of the standards around the mapping, I know that what comes next are the EDC modules. So recommendation engines can be built to support that. And then finally, another automation around his site of bringing in content from their documents that they have to put into the site holders and into the trial master file. While we were able to actually look at that information, again historically speaking, what we've seen in trials in the past be able to automatically classify it, but also check the quality level. So it really is about what we're doing here is setting up the automation functions, but also starting to check in real time and very early on somewhat of a level of pushback to say, hang on a minute, is this information coming into me of good quality? And if it's not, I want to be able to flag that immediately and in real time back to the person that gave it to me so I can have that conversation and then I can have that system helping to do that for me and I just have to basically deal with the issue of that being a problem. Beyond that, as we go into the automation, we go into the detection. So as we said, there are definitely a lot of different companies out there that do algorithms. We can all say we can do algorithms. What we're saying here is we're blending these pieces together in relation to being able to look at subjects, content or indeed site content and identifying those outlier information. And by bringing all that to the attention in the right way, you're able to start to make the connection points. So it isn't just sort of look at one piece of information and flag that because what you're going end up doing is flagging 1,000 of those, rather than it's more a case of, well, this has been identified as well as this. And because these together are very similar, that's how we start to bring up those modeling aspects and say, because of these factors and multiple factors, they are worthy of an insight that is higher in weight and important than others. And really do that at the subject level as well as at the site level, so providing those risk elements back to the user so that you can really hone in on the key ones that are important to you rather than the sort of laborious ones or maybe they just put that in data-wise incorrectly. So it really is about how to manage that, how to oversee that in the current way and present that to the user. Prediction. We are working on predictions. We do think there are some value statements to be had in that space because we do think that the earlier that you can identify information that's going to be relevant to you, the quicker you're able to prepare for, be more proactive as a user. And that's part of the oversight. Being able to oversee something you go, okay, I see there might be a trend coming this way, let me keep an eye on that is important. So if there is a trend in say being able to predict early signals around safety or around risk or even operational factors, then you want to be able to know that, historically speaking, you see these things and these are sort of typical clusters. But in this case, because of these factors, we see them running as a combination outside of that, you need to be keeping an eye on this because this could be becoming something that could be coming a major factor such as a [ protocol ] violation. So how can you track and have models with historical content that we've got to be able to put those into the library of AI tools and algorithms that have behind the scenes are running and then presenting that summarized content up to the user through the oversight to really be able to track not just, okay, I've detected something reactively because it's already existed versus I'm detecting or predicting the potential for this to exist. And we do think that allows for a much more of a proactive and oversight viewpoint, which really makes it the user hone-in on the specific needs that they've got. And then finally, as I think we said, it was a bit of a case risk, but it's not, but is okay. So I'll just flick through this. It seems to be taking a bit of time. One of those areas as an example is really around hypothesis generation. So for example, safety signals is take historical data and take indeed real-world data. And what you're taking is, what we've seen historically around clinical trial space, which is much more subject-focused and much more sort of really detailed. And then you're taking the real world wrapper around that, which says, well clusters of patients tend to have this kind of signal, and that allows you to build on that hypothesis of the potential for a signal. So that's what we're really showing there. And that allows us to then say, right, okay, we're taking upon that, that sort of evaluate the potential of this, working with the medical team to then say, all right, yes, there are these trends coming your way, what do you think? Okay. So generative AI. We do see that there are cases where there can be uses of it in the appropriate manner. So I call that cautiously because we all know everybody has basically been trying it and everybody can try ChatGPT. What you throw into it and what you get out of it is quite variable. So what we've been doing is saying, okay, that's great as a starting point of having a model that can help to understand if I as a user wants to be able to look for pieces of data in my clinical trial, I don't necessarily want to have to use all dashboards and hunt down to the next level and find this and find that. I just want to ask the question. I just want to take a couple of minutes saying, right, okay. Thank you very much, CDAS. Now tell me one of my AEs over the last month that I've had on site X. Well, to be able to tie that in, rather than dig into, okay, okay, here's a dashboard, I've now gone into the AE dashboard and they're looking at those pieces. There is the case that the generative AI can help. That has been done appropriately, as Wendy was saying, within the regulatory frameworks that you need. You can't just chuck it at out to one of those open AIs. That's not what you're allowed to do. You really have to be very careful of how you technologically-wise have your [ LLM ] model and how you make it domain-specific and how you train it. And so it really is a case of being able to have technology and an infrastructure that allows us as part of CDAS to support the user in having these kind of tools, but having them in a way that is still regulatory protected, IP protected and will provide the proper results. So to do that, it really is a case of the collection of all that data and the overset of that data into a form and structure that the generative AI can recognize. I'm not going to go into details [indiscernible] separately. But beyond that then that can be consumed in an appropriate way within a regulated model by the model so that we can then type the question in. And because it knows that data and has learned that kind of structure of data, it will be helpful and effective. And it can therefore pop up these kind of points that you're seeing in front of you. This is something it can be done, we are doing. And we think that's of value to our users to allow them to have much more of a flexible dashboard experience and oversight experience because now you've got this that can help you, it recognizes who you are when you come in, so it knows what studies you're allowed to see, it knows the data around those studies. So you ask the questions and it will find the information for you and present it back to you. And so you're having a conversation with your dashboard. At the end of the day, it's helping you get to your insight quicker, more efficiently and help you raise some interesting questions as part of your data money. Another thing that we have been working with is around the way that generative AI can actually help create other forms and other pieces of information. So if you have oversight and you're collecting the information around protocol, for example, there is a way to be able to take the content from that protocol, look at it in relation to where other protocols have run the clinical trials that you have. And what risks did you find during that historical content. And by doing so, you're able to say, right, well, here's another protocol, here's another type of content and we can therefore recognize that and as such be able to build out different plans, such as a risk plan. So that can really help to automate, but also sort of help to raise information that you may not have as part of the standard and put it into a sort of standard structure that allows you to therefore really sort of build out the operational activities that you need more quickly and efficiently. So that's another use case we see. We do see benefits in other parts of the operations of being able to bring up those kind of content. And then because you've got that, when it comes to actually during the running of the clinical trial, instead of having to go to SharePoint and find one risk plan for which study again, this look for that within a SharePoint thesis, what section of that was in the risk plan, you've got your generative AI right here and you type in what were my risks for this again? So now I've got a more again user experience that says as part of my clinical trial oversight, I can ask the questions of what was -- what can I expect in my risks to come up? And because it's there in the ChatGPT, it's going to be able to answer those questions for you. So again, we see within a structure of IT infrastructure within the correct build, it is possible to build out these kind of tools that are going to help the flow and the operations of the running of the clinical trials. Okay. Just to move on to [indiscernible] time. My closing thoughts before I hand back to Wendy is really around, can it be done? Yes, but with global regulatory authority requirements. You have to be careful. So you need to have an infrastructure behind you. And obviously, we do. We also have then the cloud technology and the digital data access rights, the appropriate access rights. If it is a really high level of security around the data, you've got to be able to understand and recognize when this comes in, what can -- I can and can't do with it. And we obviously have a lot of that experience around being able to handle that technology and data at level per the requirements of oversight and to do it promptly. And then finally, from the point of view of having these AI tools, it's great. But why I found a lot of value is in the reusability of these tools. Having built these kind of capabilities, I can use them across all my trials. I can use them as part of oversight, but also as part of other pieces of creation of other content or creation of plans or indeed questioning of plans or questioning of content in my trials. Because I've got a ChatGPT that I can train and make available and plug it into my system, I can then ask the questions and it will come back without me having to hunt down those plans or hunt down our content. I can do a lot of searches and find a lot of answers to my questions more immediately. So that's why I see these benefits as being for the clinical trial, particularly around the oversight. But Wendy, I'll hand back to you to be able to take over from here.

Great. Gary, thank you so much. Very much appreciate your insights there. So kind of bring us back around. Wanted just to put out here that the -- there's a lot of really complex big problem solving that we can do. And there's a lot of AI and automation. And at the same time, we can also like drill it down and remind everybody that this can also be a simple solution where what you're talking about is the ability to get your data from multiple various CROs, from their CTMS typically, from EDC, get it into a data ecosystem that is built for data ingestion and modeling so that the only thing you're having to worry about is analytics. And in fact, IQVIA can also provide the services so that you don't even have to work about the ingestion and modeling. It's very self-service, but we can also service it for you. So at the end of the day, you're consuming the analytics for your trial oversight. This brings value to not just to sponsors, but to CROs as well. This can actually work to help reduce the burden on the CRO around data delivery and report delivery rate and help manage the cost with the CRO around integrating your data. It provides sponsors more control and flexibility. And lastly, it -- I talked to a lot of customers about the fact that this is like a first step into a data warehouse and data and analytics capabilities. And what it can do is build a foundation for future-proofing your analytics capabilities when you do start larger trials, when you do start insourcing some of the conduct of your trials. So it also kind of brings a lot of value to the future. That's how I like to think about that. So just in closing, I wanted to give everybody a chance before we move into Q&A. This is a QR code that just takes you right to our system that allows you to request further information or a demo. And these are going to pretty much lead to a conversation with myself or Gary. So please, this will be on the next few slides. We are happy to talk about pilots and proof of concepts. We have really nice pilot program for sponsors that want to see what this looks like with potentially just one trial. So happy to have those kind of conversations with everybody. Really appreciate your attendance and really appreciate, Jeanne and her support. And Jeanne, I'm going to hand it back to you to kind of open us up for Q&A.

Thank you so much, Wendy. And Gary, thank you as well. Both of you, such great insight. And this is the future and we are living the future now. So all really relevant and important insight that you've brought up so far today. And on that note, I've got some really great questions starting to populate in from the audience. So they are just as engaged, and I'd like to thank them for their engagement in today's presentation. For anyone who may have logged in a bit late, you can submit a question you have for Wendy or Gary or both by submitting it through the Q&A box that can be found right in the video player. Okay. So leading off into our questions, our first question today -- all right, this is really good. This is really great. How long does it actually take to implement CDAS for a sponsor working with at least 5 CROs? So they're busy and multiple other vendors.

Yes, I can take that. I love that question because you have 5 CROs. It's not uncommon. So I would say what our experience has shown us is this could take anywhere from 3 to 6 months to implement. And our goal is always, we have these implementation plans and I talk a lot about crawl, walk, run, fly. So it's like let's implement for a CRO for a particular study or maybe you've got different aspects of the trial outsourced. You've got 2 CROs on this study. You've got EDC on that study that you want to bring in. That is something that we can spin up the infrastructure, help with the standardization of data from a couple of CROs and the EDC and get you analytics, soup to nuts in 3 months. Then you start adding on to that. Let's bring another CRO into the mix. Let's bring IRT into the mix. So I'm not -- it's not surprising, it's a really variable answer, but there are ways to get to value very quickly.

It's not a sprint, it's a marathon, right?

Well, we like to think of it as a very easy marathon where IQVIA will do most of the running for you.

I like that. So how can confidentiality be guaranteed?

Is that in the context of AI? Like in confidentiality? And I would assume like data privacy and security. Gary, do you want to take that one?

Yes. From the point of view of the systems that we use, so we don't use the modeling, for example, externally. We bring it in the house and we work with it that way. So there are levels of security that can be identified with the systems. And at the top level, there's certainly different ways of needing to get access to those. But also then by bringing the models in, we are controlling for how the models are developed and built without anything externally being learned from them. So it really is about how you are keeping them around your arms, as it were, so that you are not allowing it to go just outside to an external service directly. So what we end up doing is basically working carefully to see what the latest is, bring it down into our systems and learn that way within our controlled on-prem, or alternatively, we do have relationships where we can present in the cloud models in a way that are protected at that top level of security.

Okay, great. I will be coming back to you, Gary. I know we have a couple of other questions specific to AI sensitivity and all. But I have another question just to keep Wendy engaged here too. Can CDAS be used as a communication tool in terms of like team issues, identifying and communicating back and forth? Can this be used to help them take action, I should say?

Yes, absolutely 100% and in multiple different ways. So there's one of the tools that we have that we talked about in one of the animated slides is the data review module. Its sole purpose is consolidated collaborative data review and issue communication. So there's in-built functionality for cross-functional resources to be managing issues together. And then from the analytics perspective, there's the idea that we are moving towards within analytics commenting and that within analytics commenting actually is part of that consolidated shared issue management communication system. So whether you're in the data review module or in the analytics module, you're able to identify, manage, communicate and resolve issues without a lot of redundant issues and especially without a lot of queries, redundant queries that end up back at the CRO or all the way back to the site. So great question and a 100% yes.

So to follow-up, then let's bring Gary back in. Talking about what you just said, can you -- Gary, do you believe AI can be used to help with this protocol development then of trials? Can a protocol be fed or communicated into the AI system to help predict with any issues based on the protocol?

Yes. Actually, IQVIA has been working on looking at the protocol. It's taken content from other sources, let's say, around our experiences with patients and with sites. And it helps to develop algorithms that can actually look at the protocol and identify site burden and patient burden around that protocol. That helps to really trigger right upfront the likelihood of potential of issues, whether it be protocol violations or deviations or issues around inclusion, exclusion criteria, which really helps with the sort of the -- both the recruitment aspects. So it really is upfront being able to help to plan for that. So it really is a protocol analyzer. And in doing that, that then can feed what we see as the potential for issues to feed downstream into the monitoring of those issues and back to then the predictive capability of saying, right, because I see this as how historically speaking, these are the kind of the key things that could be in sort of the web map of spider's web you were. These are the key things. These are the high alert ones that then feeds down to our systems. And what we're working on right now is to be able to feed that into our models that then say, because of these, let's track against those and let's model against those for predicting the potential of that to exist. And as it gets to a certain point, then obviously, that then raises up as an insight for the user to then do something.

And Gary, you mentioned inclusion, exclusion criteria. Wendy, you mentioned the communication capabilities here. So my first question, just to talk about the communication platform. But I think I'm going to roll over, actually, Gary, you might want to allude to the sensitivities then in designing the protocol. So let me get to that one first. Is the use of AI sensitive in terms of setting up the protocol here based on side effects versus adverse effects? How do we make sure that different language is brought into this?

So being able to do the protocol design, I mean, you can look at a lot of the metadata and historical content, but you really need a strong knowledge about that. So it really is around ontology sets and NLP and having that ability to be able to recognize what is in the document, but what is the meaning of that? So just because I've got words in a sentence, how do we actually convert that into a meaning. So I do believe -- I know that we've been doing things in that space in certain areas about -- because the text is here from a scientific paper, what does that mean to the lay person and how might that fit -- pretty much we've been doing it around labels to be able to say, well, this is a scientific label. Here is a patient label. So how can I understand it? How can I interpret it? And you can't put models in for that. So in the same way with the protocol design, it is a case of being able to identify what that is scientifically, what does that mean operationally. Just because I've got that inclusion an exclusion, I need to convert that into, right, I'm including the patients for this structure, but it makes good in these kind of where they may have prohibitive medication of this. Well, that prohibitive medication could have 10 different variations of that, particularly in different languages. And you've got to bring that ontology set into standardize that. So you can look for all of the potential for that exclusion criteria and monitor against that. And that's what we would then feed downstream into the system. So the sensitivity is, don't just take it at hand of what the protocol says, but what it might have at one sentence statement, but actually bring in the coding and the mapping as part of that oversight to really bring the flavors in of what I need to therefore track is 20 different things because of one statement here, and that's what we then feed into the downstream system.

Okay. And before we wrap up, I have one just final question on that topic for Wendy. In terms of the sensitivities, again, on an international utilization standpoint here, thinking about accents, dialects, is AI sensitive here?

I think so. I think it's definitely something that the world is looking at, not just sensitive to different languages, written languages, but also to the spoken language and voice recognition technologies I think are getting better and better. Gary, I don't know, it's probably a little more in your realm to talk about like where voice recognition and potential accents impact our technologies.

Yes, and I do think it's a difficult one. I mean, you can have a call center where you basically allow to use until you talk and as such converting into text, but how accurate is it? To your point, Wendy, last few years has been getting better. I do think we have some accents we still -- we have seen in the past. A French accent or a Scottish accent can have more difficulty in converting it into the text correctly, but we are seeing improvements. What you want is a flexible infrastructure that allows you to plug that in. So in our translation team, we have the translation team go about translating obviously in different languages. They do so by plugging in whichever are the best models. So it might be for French a particular model X and for Russian a particular model Y. And it really is the case of bring all of those models, run through that, find the variations in there and learn and how that feedback mechanism that learns from it that helps you do AutoML, which is to basically choose the best model out of that. And so you can put KPIs and metrics around that, that allow you to monitor that and track for that. But yes, you would plug in the best models. And we do see them, third-party ones do have those. Again, the question is, how do I now present that private content in a secure location that allows me the regulatory compliance and regulatory issues around the Europeans AI regulations to be able to allow me to run a model in relation to where that content is, while still maintaining the privacy of that particular individual so that I can then extract the relevant information that I need to put into the oversight system. So it really is about the infrastructure and the setup of that and then the plugging in the relevant models for that use.

Thank you. Thank you very much to both of you. We are going to wrap up with today's discussion, great point, great insights. And I'd like to also thank our audience for their participation of course today. And most importantly, I'd like to thank IQVIA for making today's webcast possible. As we close out this discussion, we hope our audience will participate in a brief survey. The survey will automatically appear on your screen after the presentation has concluded, and we greatly appreciate your insight. You'll also receive an email alerting you when the webcast is available for replay. And we invite you to forward that announcement to any colleagues who may have missed today's live event, but find Gary and Wendy's discussion insightful. So thank you again everyone for joining. We'll see you next time. Bye, bye.

Thank you, Jeanne.