[Image courtesy of Adobe Stock]
Such disparities in healthcare were further highlighted when Moderna, soon to become a critical player in the vaccine race, faced a glaring revelation in late 2020. Only 24% of participants in their phase 3 study were from communities of color, despite these communities bearing the brunt of the COVID-19 pandemic. In September of the same year, the company discovered that merely 7% of trial participants were African Americans, while they represent 13% of the total population.
“This discrepancy would not have been acceptable, not only from the perspective of representing the right populations in the study but also for future vaccine adoption and uptake,” said Rohit Nambisan, CEO and co-founder of Lokavant, a clinical trial intelligence platform. But Moderna, acknowledging the issue, took steps to ensure a more inclusive and representative study of its vaccine’s efficacy. By adjusting their participant demographics to mirror the population most affected by the pandemic, they succeeded in prioritizing clinical trial equity for their COVID-19 vaccine.
Policies and regulations promoting DEI in clinical trials
In the face of this challenge, FDA is gradually rolling out new policies promoting diversity, equity and inclusion (DEI) in clinical trials. Under section 3601 of Food and Drug Omnibus Reform Act (FDORA), for instance, drug and device clinical studies must include a diversity action plan when filing certain trial documents to FDA. In April 2022, the agency published draft guidance to help optimize diversity in clinical trials.
“We observe a lot of enthusiasm among sponsors today in developing their diversity plans and complying with these guidance documents and initiatives,” said Craig Lipset, founder of Clinical Innovation Partners and former head of trial innovation at Pfizer. “But guidance is not a mandate.”
Diversity plans are fundamentally compilations of diverse strategies, which typically focus on three key areas: fostering trust, ensuring a representative patient pool and maintaining trial accessibility. “These tend to be the primary areas of investment. Then, sponsors do their best and wait for the outcomes,” said Lipset, who also serves as co-chair for the Decentralized Trials and Research Alliance.
One challenge is that FDA’s guidance document has a caveat “stipulating that if you fail to meet your plan, revert to us and we’ll assess if post-marketing surveillance is needed for specific populations,” Lipset said. “There’s almost an acknowledgment and expectation that implementing these plans is challenging, and accountability isn’t necessarily a given.”
“The vital part of our conversation here is how we can make these plans more feasible and intelligent using data,” Lipset emphasized. “We’re not just looking at dashboards to count the Caucasians we enrolled versus the others we didn’t.” Instead, the goal is to infuse intelligence, prediction and recommendations to the process. “In this way, we’re not merely throwing strategies at the problem and claiming we tried, but we are assessing which strategies make the most sense and are positioned to deliver,” he noted.
Exploring machine learning and data analytics for clinical trial diversity
A growing number of players, such as Lokavant, Medidata, IQVIA, Saama and Insitro, are employing AI to drive efficiency in clinical trials
In broad terms, Nambisan described the approach to optimizing trial diversity as a dual-focused effort. “The first involves generating an unbiased data sample for site selection and leveraging various metrics. The second component is enabling stakeholders in the study to respond to operational information in real time,” he noted.
The first element involves generating an unbiased data sample for site selection and leveraging various metrics. This process also includes integrating a variety of data sources to create a diversity index for a more holistic ranking of trial sites.
Top sites based on enrollment rate often fail to prioritize diversity goals. [Image courtesy of Lokavant]
Applying machine learning techniques such as Bayesian models and multivariate clustering, Lokavant offers explanations for its predictions. This explanatory AI empowers non-technical users to understand the procedures, thereby fostering trust and credibility in these new strategies.
Nambisan also emphasized the importance of data quality in predictive modeling and machine learning. Despite the obstacles drug developers encounter in accessing data from other sponsors, he suggested that trusted third parties who amalgamate information from diverse sources could play a vital role.
The inclusion of a real-time operational feedback loop in the trial process can swiftly correct discrepancies, encouraging proactive action and supporting diversity goals.
Predictive analytics can enable proactive problem-solving
Nambisan advocates for the potential of predictive analytics in identifying diversity challenges and forecasting upcoming obstacles. “When a site user or study staff sees that they’re unlikely to meet their enrollment or diversity goals within a specific timeframe, it spurs action,” Nambisan explained. “That’s the aim of such analytics — to inspire action that adds value to trial operations.”
Craig Lipset echoed these sentiments, highlighting that predictability is a valued characteristic in clinical trials. “If we can show that achieving diversity is predictable, it becomes a more feasible reality rather than a mere plan,” he said.
Lokavant has already deployed enrollment predictive models in several scenarios. “In fact, we’ve been able to predict years in advance when certain study teams will not meet their enrollment goals within any reasonable timeframe,” Nambisan said. “This is not about spreading fear, but rather about providing teams with necessary information.”
This capability to predict and perform scenario analysis is not only insightful but also powerful. “For instance, using a closed-loop model, we can simulate the effects of opening more sites in a country or closing non-performing ones. This allows us to test strategies in a safe and less costly environment before implementing them in the real world,” Nambisan concluded.
Filed Under: clinical trials, Data science, Drug Discovery, machine learning and AI
