Platypus: Deep Data Dives For Clinical Data Reviewer

April 2, 2018

By Allison Proffitt

April 2, 2018 | If you are creating a tool designed to give a sixth sense, of course you name it after a semiaquatic, egg-laying mammal endemic to eastern Australia.

For Krista McKee, director of data analytics at Takeda, the timing was perfect. Her son was working on a school project about Australia. She was working on a tool that would give their clinical data review and medical monitoring teams a deeper look into clinical data. It was easy to see the similarities.

“We used powerful advanced analytics on the clinical data to give the medical reviewer that power of electroreception within the data. It’s the equivalent of swimming through muddy waters with their eyes closed, but still being able to see the food,” she explained.

But first the project had to get started.

Like many companies built by acquisitions, Takeda found itself a global company made up of many different groups following their own processes and introducing inefficiencies. To address that, over the past few years, Takeda has been focusing on becoming a global organization with cohesive processes, McKee said.

Takeda launched a global review of needs within the company, and the inability to interact with study data was flagged, McKee explained. The review team interviewed more than 60 medical reviewers across therapeutic areas and functions about their data review practices and pain points, did external benchmarking, and an analysis of internal tools. “The medical function, the clinical science function needed the most help,” McKee said.

But that wasn’t the only need uncovered at Takeda, and medical review improvements “kind of fell off the global radar screen,” McKee said. Within the Oncology group, then, McKee and her colleagues decided to take a grassroots approach. “We really engaged the whole oncology clinical group to say, ok, let’s create the tool for you.”

McKee knew early on that she did not want an out-of-the box solution. Many out-of-the box tools for medical data review are, “more limited and focused on just the safety review, as opposed to inclusive of efficacy, inclusive of protocol-specific checks,” she said. And she was concerned that an out-of-the-box tool would be only a partial solution, either requiring extensive customization, or still using, “pages and pages of Excel spreadsheets.”

McKee also found that within the medical reviewers group, there was a broad range of comfort levels with technology, so she wanted to stick to a user interface that would feel familiar, and offer a tool comprehensive enough that reviewers wouldn’t revert to Excel spreadsheets. “There are some clinicians within the group that just like excel and have a hard time moving away from it,” she said. “We wanted to make sure we built something that was really designed for the user and was intuitive.”

Sum Of Its Parts

Not unlike the animal, Platypus is made up of various parts, specifically designed to meet users’ needs.  The team started with a list of about 65 general medical review questions that a reviewer would use to assess clinical trials. “We designed the system to answer them really, really efficiently at the aggregate level,” McKee said. Reviewers can then start with those answers and dig more deeply into any outliers or interesting answers, following data all the way to the patient level.

McKee worked closely with the Data Sciences Institute, a group within Takeda, to maximize how the users could interact with data. The Data Sciences Institute, in turn, tapped Deloitte to help establish data sourcing processes and dashboards across product teams and functional areas, representing performance and activities within each function.

Deloitte brought expertise to the table to “really understand and translate the needs of the medical reviewers, who this tool is built for,” explained Raveen Sharma, specialist leader at Deloitte Consulting. “We understand the vernacular, and how they describe what they want to do, then convert that into a set of requirements than can then be built out in code.”

Research Trust, part of Deloitte’s ConvergeHEALTH Miner product, is the SaaS platform solution handling all of the data. It’s a data storage model, Sharma explained. (Deloitte has been working on a broader initiative with Takeda called the R&D Data Hub, a big data platform or data lake designed to enable controlled access to the “right information at the right time with the right analytics applied for the right business groups”. Sharma calls Platypus a “very important use of that platform for a specific set of use cases.”)

In addition, the team used Tableau’s customizable, interactive visualization interface. It’s familiar to medical reviewers; it, too, has been used in other areas of Takeda. “We built [Platypus] explicitly with the user in mind,” McKee said. “We used advanced analytics to really make sure we had a powerful, user-friendly tool for this population.”

Platypus in Action

Platypus was launched with a pilot in a phase 3 study that had an immediate business case, McKee explains. The team intentionally chose a study that was already in progress and had “a reasonable amount of data,” she said. “As we built it, we’d run the data through to see how it was looking and make adjustments that way. It was a wise way to start, because it gives you maximum flexibility in how you design it.”

The pilot started in January 2017 and wrapped up in about six months. Right away the lead clinician on the study who had been doing a lot of data review reported that at least 50% of his time had been freed up by the tool, McKee recounted. The data management group also saw huge efficiencies. The kinds of errors often caught at the end of a study—data mapping errors, some findings—were flagged much sooner because data flowed through the system, she added.

Platypus lets reviewers interact with the data strategically, in a much more advanced way than a data listings. “You’re looking at hundreds if not thousands of patients at the patient level. There can be tables that are aggregated and at the summary level, but they don’t tell the story. They don’t allow you to move easily from one question to another and follow the story through to the end. This tool allows that, it allows you to do that throughout the study.”

“You have monthly access to, and strategic interaction with the data throughout the study, which is a game changer,” McKee said. “To allow for that type of viewing on the data over time, it’s just not something that’s a general practice. There’s generally a lot of listings approaches and standard timeframes… You can’t see all the data in such an efficient way, so consistently throughout as you can with Platypus,” she said.

“Essentially what we built in that pilot was something that was far superior to anything anyone had ever seen. It was extremely smart in its ability to be templated, but it also had customization element—and still does—that allows for maximum utility of the tool.”

McKee didn’t have to “sell” Takeda senior management on Platypus, primarily because the tool had proven its value in the pilot stage. “When you’re trying to do changes in a large organization, piloting is the way to go,” McKee said. “Conceiving of a pilot project, that just makes sense in the organization, and giving that a go is the way to prove that value, that you are on to something good.”

Platypus is now live in every phase 3 oncology study that is planned for readout in 2018 and 2019, McKee said. “In all the other therapeutic areas, we are working on the templates. We’ll follow the templates with an aggressive rollout.” Templates have been created for central nervous system trials; templates for vaccine research and gastrointestinal research are in the works.

“We’re still early on; we’re effectively five months out from starting that rollout. We’re working to continue to optimize the process and make it even faster,” McKee said, but she is confident that Platypus has utility across the spectrum of drug discovery at Takeda, and foresees using it earlier in the pipeline as well. “[Platypus can be used] early phase to look at clinical safety data that’s coming through, or even patient reported outcomes data that comes through. It certainly has utility both in early and late phase.”

Platypus is also working well to support Takeda’s outsourcing model, McKee said. Takeda uses PRA Health Sciences as its CRO and operational engine, and focuses its energies on drug discovery. “[Platypus] created a view on the data that would allow us to have confidence that the data are progressing the right way—or not—and allowed us to engage in a feedback loop with the CRO. So that we could, effectively, let them be the operational engine and focus more on the other activities linked to development, while still having a close connection to the data and the study,” McKee explained.

She anticipates measurable shortening of clinical trial times with Platypus. “When you think about the end of study and the time from last patient out to CSR [clinical study report], we’re anticipating that a tool like this—with its shifting of important work earlier on and throughout the study, and its allowance of strategic interaction with the data—we anticipate that it will result in a reduction of several weeks in the cycle times at the end of study between last patient out to CSR. It’s something we’re anticipating at this point that we’ll look closely at when it comes.”

Sharma says that Platypus is “generating tremendous value” now, but he is even more excited about what Platypus might enable in the future. “What’s happening is that there’s a lot of critical data that’s being amassed in the Platypus system itself. That’s going to allow the application of more smarter and advanced algorithms to pick out those outliers, pick out those interesting patients,… [and draw medical reviewers’] attention to certain anomalies or certain trends in the data,” he said. “Once you have that data available in this platform, the application of predictive analytics, cognitive, or even artificial intelligence, or AI, really becomes that much more applicable because you have the data to feed those algorithms.”

McKee agreed. “The big data platform allows use to answer the questions of today efficiently,” she said, “but also sets up Takeda for answering the questions of tomorrow, not just individual analyses but cross-study, cross-program, and even more advanced analytics.”

Editor’s Note: McKee and Sharma will be presenting their work on the R&D Data Hub and Platypus at the 2018 Bio-IT World Conference & Expo, May 15-17, in Boston as part of the Clinical Research & Translational Informatics track.