As increased diversity of patient populations who participate in clinical trials becomes more entrenched as a modern business, clinical, and regulatory requirement, new trial technologies and tools continue to emerge to help research and researchers meet these obligations. The combined impact of all these forces is helping to focus research on more specific patient populations and needs. But this impact has built up enormous complexity for clinical trial data management and analysis.
Diversity is not only driving an increase in the amount of data. It’s also creating powerful new tools to meet the challenges of intelligently gathering and analyzing these data. Clinical trial sponsors and sites must now identify not only how many but how many of which types of participants they need to ensure that the diversity in the people enrolled in their trial represents the different people in the population for their product’s intended use.
Sponsors and contract research organizations (CROs) typically select study sites and investigators based on their previous performance, including enrollment rates, captured and rendered through data analytics. But if we continue to select investigators and sites based on previous performance, and this previous performance has typically not represented the demography of that condition or target population, then we’ll continue to get what we’ve always gotten: Not enough diversity in study participants or data.
The below slide incorporates data gathered from study sites in the US and analyzed by a technology vendor for a global CRO. These sites had conducted two or more trials in the US and were selected from this vendor’s proprietary data set of more than 99,000 study sites. Site selection was therapeutic area agnostic.
The vendor then modeled the projected performance of each site based on historical clinical trial data for these sites and investigators: Not just how quickly they enrolled participants but quality measures such as how well they maintained the participants’ treatment regimen and safety and adhered to the study protocol. This first analysis, on the left side of the slide, indicates that the top five performing sites would be sites 19, 12, 11, 16, and 14.
But meeting these new requirements, by answering how many of which type, requires deeper insight into data about the participants, not just the sites. In this example, different real-world healthcare data sources were linked to participants by deidentified tokens representing their demographic, socioeconomic, and geographic circumstances. Zip codes, for example, are widely considered predictive of future health needs in economically disadvantaged rural and urban settings.
This refined approach culminates in looking at more than just enrollment rate but also at which sites are the best for meeting the diversity goals of the study–a tool that gives study site staff the ability to self-monitor, in real time, and “course correct” if necessary to make sure they’re achieving both the enrollment goals and the diversity goals. This requires several enabling technologies: Data must be harmonized across the entire study, and then localized to each specific site. Study site staff must have specific permissions to monitor their daily recruitment, enrollment, and randomization numbers at a deeper demographic level.
“As we identify investigators that traditionally haven’t been included because we’re moving away from only using preferred sites, I think industry will find a number of physicians (not investigators) without investigator experience. These physicians have participants that are diverse and of interest for a given trial, but we need to support that position by training them to be a clinical trial investigator, because it’s not an easy step,” suggests Lokavant CEO and founder Rohit Nambisan.