CROs Leading The Way In Modern-Day Clinical Data Management
By Deborah Borfitz
July 13, 2023 | Contract research organizations (CROs) increasingly find themselves on the leading edge of changes afoot in clinical data management, making it less about the policing of information and more about extracting meaningful insights from it. Among them are global CRO Bioforum, which specializes in creating efficiencies through data linkages and machine learning, and Platform Life Sciences (PLS), a CRO based in Vancouver, Canada, which focuses on adaptive clinical trials.
The ever-growing volume of data that might be collected in and for clinical research purposes—from wearables, electronic patient-reported outcome (ePRO) surveys, telehealth platforms, genomic sequencing, electronic health records (EHRs), and patient registries, to name a few—has necessitated the shift to the data science approach. It is the industry’s way of “coming out from behind the curtain” to start better engaging patients, their ultimate customer, according to Tanya du Plessis, chief data strategist and solutions officer at Bioforum.
This is business 101. “The only way to get medicines to patients quicker is to get them to the market quicker and, to make that possible, clinical trials must be engaging,” she says. Participants need to feel involved, heard, and appreciated for their role in the research study. Putting up stumbling blocks, even inadvertently, will instead ruin the experience.
It would be ideal to have all clinical research data managed in a single hub from which insights could be seamlessly drawn, continues du Plessis. But for now, the aspiration is to at least get what’s being generated by a trial (e.g., lab data, electronic case report forms, readings from medical devices) into a single place.
When it comes to real-world data, things get tricky, du Plessis says. Technology companies with wearable and telehealth platforms, who understand their role in the clinical trials arena, might readily marry up. But EHR software vendors aren’t necessarily thinking about the long-term value of standardizing data for down-the-line research purposes.
It may not take long for the tide to turn because pressure is building for data harmonization by EHR systems, she adds, noting that the U.S. Food and Drug Administration (FDA) values eSource. Trials need data that live in locations outside their control, including hospitals and doctor’s offices, but healthcare providers aren’t anxious to change the way they work or to go through the process of having data standardized given their many competing priorities including, of course, taking care of patients.
Then and Now
Back in 2008, when Bioforum was founded, clinical trials were “very linear,” du Plessis recalls. “Data geeks like me were pretty much on the [study’s] backend.” Data typically came in from one source—whatever an investigator recorded in patient files or source documents and then transcribed into an electronic data capture (EDC) system or, in some cases, sent via fax.
Whatever system was used to manage data was typically visible only to data managers and statisticians doing the data analysis, she notes. Patients were rarely asked to directly enter their own data into a diary and that information was not easily channeled back to the physician-investigator.
In those days, the role of data managers was to ensure data came in clean, complete, error-free, and, ultimately, connected, says du Plessis. Today, investigators are still completing notes on a source form that feeds into a data management system, but that information only accounts for about 40% of all trial data being collected.
“Patients are constantly sharing how they’re feeling and how they’re doing, on ePRO [systems] for example,” she continues, most often via an app on their smartphone. When they wake up in the morning, they are digitally prompted to report how they’re faring, and those responses automatically upload into a central system that simultaneously receives data from multiple other sources such as information captured during virtual study visits between investigators and patients.
The upside of collecting data electronically is that it removes the potential for transcription errors, says du Plessis, freeing data managers to focus on pulling out useful, higher-level insights about the trial itself for informed decision-making. Are study visits being completed? What are the pain points for participants? Is patient safety being properly monitored? Why is a particular drug being used to treat an adverse event? Is additional training needed?
Operating through a pandemic taught the clinical trials industry that it’s possible to move faster with more data sources, says du Plessis, although the information is still largely being managed in silos. The growing adoption of data warehouses, however, is a step in the right direction.
Warehousing data from multiple sources in a single database is key to gaining insights that would be missed looking at datasets independently. “That’s where it’s super-important to have really strong and innovative technology partners,” she says, specifically referencing Veeva Vault CDMS.
Leveraging Machine Learning
These days, machine learning (ML) should go hand in hand with data handling/cleaning and programming, du Plessis says. At Bioforum, this takes the form of an ML tool called JetConvert that seamlessly converts the data collected during clinical trials into the necessary SDTM (Study Data Tabulation Model) format that is required by the FDA when companies are submitting a new drug application to the agency.
This otherwise time-consuming conversion task was immediately identified as an untapped market opportunity for the CRO, says du Plessis. Multiple sponsors (and other CROs) are now taking advantage of the SDTM automation platform, which slashes the conversion time from the traditional eight weeks to just three weeks.
JetConvert works by running the ML model over raw data, surfacing suggestions about how to map every data field to its SDTM counterpart, she explains. The human in the loop can accept a recommendation or reject it by either manually overriding it or asking for another suggestion. In addition to the time savings, the quality of submissions improves because “the consistency is fantastic.”
The ML model was trained on over 100 studies, so it knows how to correctly map standardized datasets seen across all trials, adds du Plessis. The human experts therefore spend most of their brain power on mapping the 20% of study data that is therapeutic- or indication-specific.
Further leveraging the time- and cost-cutting potential of the ML engine remains high on the list of priorities for Bioforum, she says. One target will be better managing the aggregation and cleaning of all the proliferating source data.
Regulators are clearly signaling that industry needs to start doing things differently, says du Plessis. Most recently, the FDA released draft guidance with updated recommendations for good clinical practices (GCPs) aimed at modernizing the design and conduct of clinical trials, making them more agile without compromising data integrity or participant protections. The document, adopted from the International Council for Harmonization’s (ICH) recently updated E6(R3) draft guideline, modernizes GCP recommendations encouraging the use of fit-for-purpose innovative digital health technologies such as wearable sensors.
Change is coming over the next two to three years, du Plessis predicts, most notably in terms of using data to make decisions and manage risk. Data will be coming in faster than ever, and ML will be tapped to reveal patterns and trends from that deluge—leading to studies closing sooner and a better trial experience for participants.
As a CRO supporting the clinical trial process, du Plessis says, “It’s a thrill and so rewarding to know we’re playing a part in a very intricate industry that, in some cases, is [literally] giving people a new chance at life.” She asks that her industry colleagues “challenge the norms” at their organizations by introducing new ideas and technologies that facilitate that life-saving potential.
Many Trials In One
Equitable healthcare for all is the ultimate mission of Platform Life Sciences through the vehicle of well-executed adaptive clinical trials, according to Michael Lambert, head of partner relations. To that end, the organization has been building and expanding its network of partnerships in low- and middle-income countries (LMICs).
PLS is a pandemic-born company that ran the TOGETHER Trial, the largest placebo-controlled COVID-19 study that rapidly screened 30,000 patients and enrolled 12,000 in community-based settings, he says. It was designed to simultaneously evaluate repurposed medicines in LMICs against a single control group and had a pre-specified data analysis plan. The study received a Trial of the Year Award from the Society for Clinical Trials in San Diego, enabling it to grow a global network of trials.
From a data management standpoint, the TOGETHER Trial required dropping information about a large population and 11 different drugs into the same database. “It was a huge win for our organization... and across the board for the adaptive platform,” says Lambert, who joined PLS last year.
PLS Founder and CEO Edward J. Mills, Ph.D., along with McMaster University Assistant Professor Jay J.H. Park, Ph.D., co-wrote “Introduction to Adaptive Trial Designs and Master Protocols” (Cambridge University Press), based on experiences gained in the TOGETHER Trial, he points out. PLS can be thought of as a “master protocol author” that subsequently executes on it in the form of an adaptive or platform trial.
The need for speed is non-negotiable with platform trials, he adds. “You don’t have time to wait two to three months for a change order to go through or work with [data] migrations from one system to the next.”
Adaptive platform trials have built-in efficiencies that reduce the cost of conducting multiple studies by using one sustained platform for many years as interventions get flexibly added and dropped, explains Lambert, noting that the TOGETHER Trial is still running with more than 12,000 patients. The size and complexity of the study was too much for their existing system, so, as a collaborative, the decision was made to switch to Veeva Vault CDMS.
The technology suite delivers EDC, a platform for aggregating and harmonizing clinical data sources, and a visualization tool for randomization and drug management. The fact that all those features work together is what sets it apart from some of the other clinical data management systems on the market, says Lambert.
PLS plans to start a second platform trial later this year, a randomized phase 3 study called REVIVE for long COVID. The study will look at the safety and efficacy of repurposed therapies and novel drug therapies across the globe, he reports. “We are actively speaking with many sponsors now who are very excited about platform trials and the possibilities in the works,” but will also likely need time to adjust to new approaches for conducting a clinical trial.
Veeva La Difference
Lambert’s decision to bring in Veeva technology is related to how the database is set up and how post-production changes are handled. “I have used many different systems where it was somewhat painful to move from one protocol version to the next—if you needed to update an eCRF, for example, you’d have to take your trial down and have everything run through a maintenance window.”
That is unnecessary with Veeva’s CDMS, Lambert says. “You don’t have reliance on custom functions that take an extra layer of expertise to program.”
Lambert additionally notes that Veeva CDB (clinical database), a tool that accepts multiple data streams or sources in real-time, can be queried from an electronic dashboard “versus having to dive into the database itself and navigate to a form.” It’s a snap to get to datapoints inside FDA, ePRO, imaging, and laboratory datasets.
Veeva has its own clinical language so users can quickly do their querying and resolve issues with no wait time, says Lambert. The alternative can be a drain on time and resources—notably, that of in-house programmers needed to build and regularly update databases to accommodate reports needed by clinical monitors or physicians.