New Legislation is Driving Diversity in Clinical Research
Industry leaders discuss how technology can help companies adapt to the regulatory changes and accomplish their DE&I goals.
The scientific method is most effective when testing a hypothesis on data that best represents the population. While this might be a well-known requirement for ensuring sound science, the goal of true patient representation across clinical trials remains stubbornly elusive. Moreover, attempts at obtaining diversity in clinical trials have been half-hearted at best, often relegated to a secondary priority. Now, the Food and Drug Omnibus Reform Act (FDORA) will demand prioritization of representation in clinical trials, and companies that fall short will face repercussions.
So how can already heavily burdened companies take on this new challenge and successfully diversify their trials? Redoubling their commitment to advanced technology can help companies alleviate some of the expected financial and operational stressors, as well as help them accelerate and even increase efficiency in recruiting patients, including those who have been historically underrepresented.
I spoke with three industry leaders who are already leveraging technology to increase population representation in clinical trials. Jennifer Byrne is CEO of Javara, which provides clinical research access for patients at the point of care; Irfan Khan, MD, is CEO of Circuit Clinical, a community-based research organization; and Marie Rosenfeld is formerly senior vice president, head of clinical operations at Astellas Pharma. These seasoned professionals shared their views on how technology is key to cost-efficiently accomplishing diversity, equity, and inclusion (DE&I) goals.
Rohit Nambisan: Do you think FDORA will make a real impact on diversifying clinical research?
Marie Rosenfeld: At most companies, the top priorities when considering clinical trial design and conduct are focused on delivering scientifically valid data as quickly as possible while minimizing costs. The goal has always been to make decisions with the right data at the right time to best estimate the likely success of the product and inform future internal investment and the return on that investment.Diversity is often only discussed in later stages of development (Phase IIb/III clinical trials), and not at the top of a long list of priorities.
Would a company be willing to make a technology investment if the sole factor was to increase diversity in a clinical trial? Companies are already advertising, partnering with advocacy groups, and tracking a variety of data about patient recruitment at levels that we’ve never done before. While it’s difficult to predict whether additional, substantial investment in technology dedicated to helping meet DE&I targets would yield fruit, it’s possible. As with all investments, the return needs to outweigh the cost. I think the probability of adoption is higher when the technology can be applied across a company’s portfolio and supports other efforts. FDORA will certainly enhance the likelihood as companies will be incentivized to take this important step forward.
Irfan Khan: While regulations alone are unlikely to achieve desired solutions on their own, they work best when they outline what evolution is expected. By that bar, FDORA promises real change on an accelerated timeline. It isn’t that sponsors and sites haven’t known representation in research is a real and pressing concern—but this needed push provides the motivation to fix an inequity in ways that will pay dividends for all stakeholders.
It doesn’t take much evolution to graduate to a revolution. The movement of expanding clinical research into non-academic community healthcare pioneered by teams like Circuit, Javara, and many others provides a simple way to positively impact representation in research. By choosing to partner with health systems and physician groups that serve communities of color, we have been able to empower participation at rates more representative of the expected population. That finding is not unique to my team and suggests that when patients understand and trust the source of the information on clinical trial participation, their interest and support for participation improves dramatically.
Jennifer Byrne: FDORA is unmatched in that it addresses two important realities facing our industry: diversity and modernization. When drugs are approved, we assume—falsely at times—that they have been accurately studied in the patients that will be using them. Unfortunately, that is not always the case. If the clinical trial population does not mirror the patient population it is intended to serve, we lack a clear understanding of how it will work when prescribed.
Consider, for example, Javara’s investigator-initiated, decentralized trial focused on surveillance—identifying symptoms and how they spread. Launched at the beginning of the COVID-19 pandemic (April 2020), within six weeks, we had 22,000 people enrolled, reporting daily symptomatology. What makes this number significant? Less than 1% were Hispanic, while representatively, that number should have been 12%. African-Americans comprised only 8%; they should have accounted for nearly 30%. The lack of population representation in this example was detrimental to the data and treatment options obtained through the study.
This issue is very real across clinical research and firm regulatory guidance is without doubt a key part of that solution. FDORA is different because it applies directly to all of us. I have no doubt about its ability to greatly impact diversity in clinical research.
Nambisan: How can technology, and specifically artificial intelligence (AI), help solve such a vexing issue?
Rosenfeld: I don’t think we have figured this out yet. AI has the potential to revolutionize how we think about clinical trial execution. I think improved DE&I in clinical trial settings will be a natural outcome of improved patient identification, utilization of digital technology, and strategic efforts to enable patient access to trials. But to see the full potential of AI in research, the entire industry needs to track closely with the modernization of healthcare. I think one of the things COVID taught us is that healthcare is still very personal and not every patient or physician wants to interact with technology the same way. This will continue to be a significant hurdle to more innovative applications of AI and technology in research.
Khan: First, it is important to define AI by today’s standards. For instance, are we talking machine learning algorithms that can do more and more with unstructured deep datasets like EMRs (electronic medical records)? Or are we thinking about generative AI tools that could potentially improve how we communicate messages across more and more specific audiences? I’m tech-adoptive by disposition and see much to love in gen-AI’s “100-foot wave.’” But I’m skeptical that any but the most niche solutions are ready for prime time when it comes to impacting something as difficult and bespoke as matching a patient to a trial and effectively engaging them. Instead, I think their greatest promise in today’s early phase is workflow automation for overworked teams trying to parallel process reams of information to understand, “who would truly benefit from this conversation about this clinical trial?” If we solve well for that phase and can improve research teams’ quality of life and quality of work [freeing them from more mundane tasks to manage a greater number of patients], then I think that’s real and relatively near impact.
Byrne: A new sector is emerging as research and care converge. I strongly believe this new space will enable us to work more effectively with the immense quantity of data available through health systems and use AI optimizations to reach those known patient populations. We start with such a wide funnel; this empowers us to get it down to where we can predict those known patients. I think AI can help us address the ethical, moral, and public health responsibility we must reach these patients with. Our inability to do so thus far is at the heart of the problem today.
A huge percentage of patients do not trust pharma or our healthcare system in general. By making a genuine, targeted effort to reach those patients, we can start building trust.
I also see AI [helping] to democratize research. It enables us to quickly look at the unmet needs of a patient population and then match those needs to a sponsor’s portfolio. It can also help ensure the fidelity of messaging and community outreach. At a recent community event, I found myself in conversation with the head of the local Black Chamber of Commerce. He exuberantly said, “Javara works with my doctor!” He continued saying how much he loved the fact that we seek individuals like him to participate in our programs. In the past, members of this community would likely have shared skepticism toward clinical research. To hear this individual express appreciation for our focus was a testament to the reliability deployed in our messaging. AI can—and is—playing a role in new processes and technologies that ensure clinical trial opportunities are offered to all patients.
But while technologies are important and make it easier to find and address certain patient populations, it is still vital that we focus on reaching patients at the point of care. It’s high-touch and how people become comfortable participating in research.
Nambisan: What are some of the hurdles to companies leveraging modern technology to solve this access issue?
Rosenfeld: One of the challenges is around technology implementation. Sponsors need to enable sites and patients to utilize their own technology. The pharmaceutical industry has encountered challenges time and again from trying to force technology solutions on sites and patients. Companies build infrastructure for a perfect data flow that will theoretically save time and money, and then we try to force the solution on the sites and patients. Increasingly, this strategy results in costly workarounds for either the site or sponsor to map or fit the data to how it “needs to be” for the sponsor technology to work.
The key question has been, how to get the information sponsors need from site and patient data sources that already exist, supplementing only that data which is specific to the trial? AI has the potential to significantly change this paradigm. Sites say they are overburdened with process and technology. Trying to manage all the sponsor technology requirements not only steals attention away from patients, but discourages physicians from participating in research—ultimately, compounding the diversity problem.
Khan: The biggest challenges are data access, data privacy, data security, and data respect. How do we leverage large healthcare datasets to engage patients at scale with a level of precision-matching previously impossible in brick-and-mortar sites using traditional radio and social media ads, yet still ensure patient privacy? Healthcare systems—particularly those that serve communities of color—are understandably cautious about granting access and under what circumstances. It comes down to trust and respect, as all partnerships eventually do. This caution is one of the great limiters for pure platform companies—the trust cycle is slower than the information and sales cycles.
Yet, scaled healthcare systems are now trying to monetize their datasets to technology vendors looking to train algorithms and to population health solution providers. Their view is that the “locked fortress” mentality has been holding innovation on behalf of patients back. At the same time, clinical trial sponsors seem more willing to accommodate trial sites and healthcare systems. Rightly so, “growth at any cost” has stopped being the focus while sustainable growth is receiving its due. The big challenge is determining how to partner with the places that patients trust, and who will fund that work given healthcare timelines are very different from those of institutional investors.
Byrne: The biggest light bulb that has gone off recently in our organization is that prescribed technology strategies can be detrimental. Every situation is different. The technology we choose depends on the unique region and circumstance. For example, we have a provider partner with about 2.5 million patients and 2,500 physicians; in this situation, our clinical decision tool enables us to successfully enroll half of the patients for this trial. That’s almost unheard of. People like to say that physicians don’t refer their patients to trials, but they do. When we leverage technology to integrate clinical trials seamlessly into physicians’ daily workflows, that’s what happens.
Before we had data sources, I drove from site to site to find potential investigators. AI can be transformative for us in efficiently identifying the right fit for a trial. With modern predictive analytics, can we really lean in and look across those 2,500 physicians within our system? Can we dig deeper and get better at predicting the best sites for a study? Yes, and doing so would not only help transform trial execution but would also move the needle with diversity and population representation as well.
As with most technological processes, AI depends on data inputs. If we don’t have enough data—speaking the same language—going into the model, we will not get high-quality outputs.
One advantage we have seen through the Javara model is the ability to leverage the amount of standardized data accessible through our healthcare partners. If we didn’t have visibility into their EMR, we would face a much larger hurdle in implementing some of these valuable technologies.
Nambisan: Where will the industry be at the end of 2024? What progress will be made in five years?
Rosenfeld: By the end of 2024 [and beyond], the industry will see significant investment in technology and AI. Better diversity in clinical trials will be a natural outcome of this investment. If we can develop solutions that improve access to patients and data, then we will naturally enhance inclusivity. Without an easy way to access a trial site or even to access information about a potential trial, we cannot move forward. Currently, sponsors rely on third-party patient screeners or multiple, stand-alone and costly databases of information. If we can leverage technology to get trial information to relevant patients and make it easier for patients to participate in those trials, then diversity will follow.
Over the next five years, I hope to see the corresponding changes to the regulations around clinical trial conduct. With all the advances in technology, we should be able to leverage the same types of solutions to better protect the rights of patients and ensure ethical and valid research. It’s hard to reconcile that we can leverage technology in novel ways, but we still get hung up on how to properly complete an Form FDA 1572.
Khan: We will see teams sharing their successes and their struggles more openly and more proactively. If one innovation team, sponsor, or site solves a problem in this area, there is no downside to sharing that knowledge, and plenty of collective good to be accomplished. This collaboration has already enjoyed a strong track record at conferences and informal communications, but I think we’ll see more people share prospective data by the end of the year on specific approaches they pioneered and won or learned from. I’m excited to share perspectives and evolve as a community.
Byrne: The DE&I pioneers in our space lead fearlessly and by example, challenging—and empowering—us to think critically about this issue. AI is a powerful tool to support DE&I advancement, but it must be leveraged appropriately. Success depends heavily on how the product is built; factors such as who, what, when, where, and how all play a vital role in determining impact. Bias can and has been historically and inadvertently built into many systems, and AI is no exception. At Javara, we take a critical approach to evaluating the use and efficacy of such products. We know the only way to ensure equality of outcome is to ensure equality within the teams and organizations creating, testing, and making final decisions about products such as AI, and, unfortunately, this is not often the case.
The pulse oximeter, for example, is a great piece of technology, but it was not properly tested on melanated skin. Disparities in care based on race have emerged as a result. Similarly, built-in bias makes it hard for patients to navigate automated phone systems, depending, for example, on dialect or accent.
These are just two examples of the potentially detrimental impact of overlooked bias. Mindfulness is key to leveraging these tools for equity in clinical research. I believe some of the biggest concerns currently stem from human complacency, biases, and processes and can be exacerbated through AI. Those who have rushed to implement AI solutions without critical reflection about the potential for bias will likely be forced to reevaluate the role of these technologies in the near future.
Rohit Nambisan is CEO and Co-Founder of Lokavant
