It is my pleasure to kick off our first track of the day. This is Dimitrius Zambas and Doug Chance and they are talking about implementing the next gen EDC cd MS platform advancing data analysis capabilities.
So welcome, So good afternoon. So I'm going to interview Demetrius, who works for a coverany I used to work for, and we're going to talk about the UDMS. So, Demetrius, first off, what is DMS.
Good afternon, Well, we made it up. It stands for Unified Data Management Solution or system, and the focus was really to do something in the sense of DEBT to capture data simulation deta consumption that wasn't entirely focused on the EDC side. At this point, about twenty five percent of our data comes from the DC. The other seventy five percent comes from third party sources, you know e pros, echoas,
central Labs, imaging centers and so on. So rather than repeat the same cycle of focusing on the sexiest, most glittery EDC when we set up the team to assess tools, we put a heavier weighting on the back end capabilities to make sure that it was something that was a little more a little more points for the future, if you will. On the front end, we actually spoke to some sites and got some feedback there. On the front end, the only matter is simple, simple, simple, Let the ciras
do their SDV. Let the sites have very simple, straightforward dead entry, simple straightforward ways to capture their normal ranges for their local labs. You know, the demand is is much more straightforward, I think, on the front end than it.
Is in the back end.
So what would you say the benefit of u DMS is over the conventional data management systems that we're.
All used to.
Well, again, what's it that imaginement system? I mean to me that imagement system was contral.
You're not showing your age at all.
No, no, no, But you have to step back and say, what is it?
Why are we collecting all this stuff? Well, what's the intention?
Right, It's not just for because we want it, although in some cases we have clinices that just want to collect more because they just want it. Fundamentally, it's to get to the point of having a fit for purpose data set to support a submission for an approval for
a new therapy. So working kind of from that point backwards, whether it's whether it's having standards that are completely end to end, and a lot of people say end to end, but end to end meaning the landing place in STTM for every single data point we collect is the before we've collected it. That irritates some people because it takes
a little bit of work during the setup. But the difference, you know, when I joined Feizer, we were at the very bottom of the of the industry benchmarks for last patient to submission, you know, the last pibolical trial, last patient, last visit to submission.
We were dead last.
I was afraid you were going to say dead lasting data management because I headed updated.
Well, data management had been dissolved.
It was fully outsourced, right, and it wasn't for the fact that it was outsourced. Sometimes you have to make sure you have the right recipe for the right oven.
In this case, I think it was right or wrong oven.
It was definitely the wrong recipe and every study in the same program was done in ten different standards. Well, then it took to our statistical programming group half a year to get it all normalized so they could, you know, do a proper summary of safety and efficacy. We're now consistently, if not first but in the top quartile. And yes, it's because it's better medical writing, it's better statistical programming,
better statistical programming environment. But I truly believe the single biggest factor is that every single study is leveraging the same superset of standards so that when it comes out, it's ready for analysis the next day, and it's not. It doesn't mean we don't add standards or change standards. It means we do it before the study starts.
Not after.
So it's been a lot of work. I mean, tell me, why is this a big game changer for.
Five Well, we so we had I think forty people in an assessment team, and we did the usual We send people out to conferences to s write down every possible player and then selected a subset of about a dozen to do a general you know, request for information response, and then from there selected a smaller subset to do a full RFP. And in that process we made sure it's usually more of a of a d M LED
activity picking an e DC. But because of what this was, it was much broader with ECO representation from clinical, from study management, from monitoring and so on, and everybody got everybody got to identify criteria that were important to that part of the organization, and then we waited them based on what we thought was most most critical to the company from a platform perspective, like like the like the ability to handle back end data loads and things like that.
So I've done this.
I had the benefit of doing of working on the first industry implementation test transfer of an e DC with my expos Dave de Torre, who's somewhere back over there, face forwards inform into sharing power right, and then I had the benefit of kind of redoing that at Mark and then when I was at Nevardist I left before I ex implement did, but we had started implementation of medidata rave there. So having done this a few times kind of you know, seeing the good, the bad, and
the ugly. When we formed this team, I think really focused on it being more cross functional, really focusing and all the different all the different components of the puzzle, not just the DC piece. You know, it's very easy to get to get distracted by the shiny, glittery lights in a new DC when you go to someone and say, hey, I wanna I want to standards super set and I want to make sure that my my c disc is all adequately defined.
Before I start to study. You know, your executives fall asleep before you're done talking.
They want to hear that it's somehow gonna be you know, solved world hunger for clinical trials. But in this particular case, we got enough with enough momentum behind it cross functionally to really really not focus on the glittery parts, but the fundamental platform parts. And no, no data what do we call you know, no batch loads? That that is
live from collection to consumption. It streams, it's live, and no one thinks that's important until you run a COVID vaccine study with forty six thousand patients and you're waiting for ninety cases and you wake up at five in the morning, You get your coffee and you go over to your desk and you turn it on to see what the data status is from the night before before you fall asleep.
It's batch loads, no muss.
So you put a lot of work into this. I should have asked this question earlier, But why what made you? What led you to this place where you developed it?
Well, we had a list of things we really wanted, kind of everything from a nice to have to we think it's an absolute must have. So for instance, the live the live data streaming. Not everyone agreed, but some of us thought it was a must have, especially those that went through some recent critical experiences, less hops, less instances of the data again, things that when you when you talk to a senior executive about prioritizing something about
a tool, it doesn't sound important. At a prior organization I worked with that started with the letter M. Someone decided to take data down the path to SDTM and too adam in parallel paths.
It was one of the most critical submissions in the company's history.
The FDA came back and gave us a month to prove to them that the SDTM and the ATOM were equivalent. It was the most horrific, stressful month of my life. I would never want it. So now we have these kind of caveats, there's no parallel data flows, no this, no that, just to keep ourselves out of that kind of trouble.
Uh. And it does. It doesn't mean we don't we can't be innovative about it.
You just can't go will and Lilly taking data down one hundred different paths, so wanting to consume it, meaning you know, drive all the reporting from the layer that's streaming live from the front end. That was an absolute prerequisite requirement.
So Sabina is such a bad experience. Never again, is what drove you? Drove you in that direction. So I'm guessing that not everything went perfect. So tell us about some challenges, you know, what lessons did you learn and implementing a major overhaul?
Yeah, well, I mean, not everything goes perfect when you're implementing something completely off the shelf, as it is because of.
The amount of integration.
Yeah, and when you're a company, the size of hours, the amount of integration is out of this world, even down to sending eesa messages to argus they have to be a certain you know, the level of specific specificity and the amount of specs that the safety team has of how they want that data and how they want the updates to that data. You wouldn't believe you think
it's one requirement, It's like one hundred requirements, right. So definitely the integration where we thought some things would be more straightforward, we're not.
The the actual getting agreement and alignment.
We had a great partner in this, but the partner you know, was part of the RFPD except all those specifications. But a SPECT that's written by a data manager or study manager or CRA written by read by an IT person doesn't always mean the same thing. So we had a lot of back and forth to get on the same page of what things meant.
And then given the scale that.
We are, the number of acquisitions we've had, we've literally had we have studies and everything.
We have studies in data labs. Actually I think the last one just shut down.
But until recently, we had a study in data apps, Oracle, clinical metadata, rave Viva, from acquisition, you know, everything, everything. So this particular implementation we are migrating. It's going to be the it's this biggest migration in the one hundred and twenty studies from Legacy Pfizer and I think thirty studies from Legacy, about one hundred and fifty studies are going to get migrated. And we don't mean build a new study and migrate the content. Lily did that last
year with about twenty thirty studies. We're actually creating the data models in the new system with a migration tool and then pulling the data over and getting all the study teams on board. One hundred and thirty of them with the best week and months for their study to basically pause.
For us to do this. We're almost We're almost there.
I think the first studies, the first migrated studies, are intesting right now. The system went live, the first study went live. I think we have our first patient in the study. No, not yet, they keep It's what almost what the study team felt it was important to accelerated from mid September to meet August for the build was still waiting for the first patient. At least you're not crying, no, no, no, yeah, yeah, yeah.
But it's one thing to go live. It's another thing. One thing to build a study and then to present test data.
It's another thing to have a site actually starting putting patients in it.
So, I mean it's kind of a wrap up. Why are you excited about this?
What?
Why should we be excited for you?
Well, it's new. I mean you're not. You're not.
You can't go and say I'm leading this space or my team is the most advanced in this space. If you're just getting another another same off the sheelf tool as everybody else. Uh here was an opper and and the big boys. There was a time when they were smaller, and you could you could influence that.
Just even with face.
Forward before they were acquired by a bigger company. But I think here we had the opportunity to really sit down and say, what's what do the sites want, what do the cis want, what do the clinicians want, what did that managers want? And found a partner that was crazy enough to jump in with us to do it. So I'm truly excited that we're going to have the most advanced unified EDC data management environment in the industry.
And I'm sure we'll have bugs as we start to execute the first studies, there's no doubt, but everything from how it's set up, to how the third party data is assimilated, to how that it flows, to reporting and to safety is it's optimized across all them instead of you know, kind of pick one or two instead of all of them.
So in the last minute and twenty some odd seconds, what's next, What's what's the next version or the next enhancement or the next innovation.
Well, we have some we already have planned for. So this is version two of red kept Cloud that's going in now. Version one was already available to the public before we've already got specs for version three, which my team wants to you know, screaming that we haven't even finished version two, but there's critical things we want to
have there in version three. And then based another work we're doing, whether it's the setup or the signal detection leveraging a gentic AI that I think we're talking about the tomorrow in another session is very exciting, especially when it's live and you can kind of prevent issues from proliferating, you know, running running a signal detection for incential monitoring
and once a quarter which many companies do. Do you know in a vaccine study once a quarter roment is finished you or even if you even if it's not finished, if you have thousands of patients through and you realize they're interproperly reporting AEES or whatever the case is. So I think think starting with this being live, the third part of data being loaded in a romatic way and having single detection running on it live, it's got to be the it's it's got to be how we have to think to be strius.
It was great talking with you this afternoon. Thank you very much.
Yeah, thank you