Paving the Way for Virtual Control Groups

A status report on pioneering efforts by Charles River to replace control animals with virtual controls in nonclinical safety studies

A pilot project launched last year by Charles River with the support of several of its pharmaceutical partners is helping to reshape how nonclinical studies might look like in the future. The pioneering work is leveraging years of legacy data, loosely defined as groups of data points created from historical control data from past studies and seeing how these “virtual” controls stand up to their “live” animal counterparts in toxicology studies.

While it is still early days, experts believe virtual control groups (VCGs) could lead to reductions in the numbers of animals used in toxicology studies. VCGs are also seen as a way to potentially lower costs, freeing up capacity not utilized by control animals, and increasing statistical power of studies, with the possibility of leading to more accurate and reliable results. This can be especially beneficial in studies where control group variability is high, as it allows for a more precise comparison.

Thus far, the pilot data from pharmaceutical companies, and Charles River’s own internal studies, look promising, says Dr. Will O’Neill, DVM, MPH, a veterinary pathologist at Charles River and one of the leaders exploring the feasibility of VCGs as a replacement for rat and large animal controls. O’Neill said they have observed no differences in major study conclusions in the VCG arm compared to a conventional animal control group, and they have been able to identify, for the most part, the appropriate target organs in anatomic pathology that matched the original study, as well as clinical pathology changes.

O’Neill said they did see some variability between the original report and the analyses with virtual control groups, but the high-level conclusions of all the studies have remained the same. “Our next step is really to understand that noise, that variability, and what the impacts might be, for example, as a compound moves into clinical testing,” he said. “For instance, are we really seeing differences that would impact what is monitored for human safety?”  

Virtual control groups as a benchmark in tox studies

VCGs contain curated matched control data collected from previous studies. The control arm of drug study does not receive the experimental treatment being tested, allowing researchers to compare the results of the treatment against a baseline to accurately assess its effectiveness. In other words, the control acts as a benchmark to see if the intervention is truly causing any changes.

While the use of VCGs is an established concept in human clinical research, particularly in studies where establishing a traditional control group is difficult or ethically challenging, such as the testing of rare or ultra rare disease, they are rarely used in animal studies. That scenario began changing in recent years, due in part to the push to reduce the industry’s reliance on animals and the recognition that years of research had accumulated a repository of high-quality data.  The FDA Modernization Act, a series of bills aimed at modernizing the agency’s operations and technology, long with a push toward developing New Approach Methodologies has also contributed to interest in preclinical VCGs.

Charles River, one of the largest contract research organizations globally and a powerhouse in Safety Assessment studies, is one of the companies pioneering the use of VCGs in animal studies. According to Dr Steven Bulera, PhD, DABT (Charles River Chief Scientific Officer – Discovery and Safety Assessment), “Charles River is seen by the industry and various consortium as a leader in this area. Will O’Neill is recognized as a subject matter expert and has presented and continues to be asked to present at various pharmaceutical companies and scientific societies. Charles River will continue to support this important and transformational effort”. 

Charles River is working on this effort in partnership with Sanofi, and collaborating with other large pharmaceutical companies to help modernize drug development and reduce animal use. But Charles River is not alone in this space. In Europe, the VICT3R project, a public-private partnership funded by the Innovative Health Initiative based in Brussels, is also looking into the feasibility of VCGs. The four-year effort led by Bayer AG and Pompeu Fabra University in Barcelona includes 25 industry leaders. 

O’Neill said right now Charles River’s VCG work is focused on rats and a large animal species. “If the method works as we expect, we could be looking at up to 25% reduction in some studies when we can completely replace of the concurrent control group and slightly less in studies conducted with partial replacement. In the not-so-distant future, we may also be looking at synthetic control groups, created by advanced computer modelling and possibly generative artificial intelligence.”

Where are virtual control groups a good fit?

When Charles River’s launched its VCG project, it tested the approach out informally to collect data and assess initial results. These non-Good Laboratory Practices (GLP) exercises resulted in additional matching criteria and discovered areas that require improvement. O’Neill said their first studies for commercial clients will also be in a non-GLP setting, which by design is less regulated in order to allow greater flexibility in how studies are executed. Initial studies will also likely be a hybrid design, which means that VCGs will work alongside live concurrent control animals in a study to build confidence in the process and provide regulators a chance to see some of these data packages. Over time, they expect the number of animals to be reduced, expand to full GLP validation, support new species, new study types, and examine the potential for complete replacement by VCGs in some study designs.

“Right now, our virtual control group database is comprised of multiple, complex datasets from multiple sources. We’re also scanning in glass slides for use in digital pathology,” he said. “All of that requires GLP validation, which is possible but may take years. We don’t want to attempt that full validation until we have our final product fully designed and ready to go. It will be much easier to start on non-GLP studies, but from our research and our exercises so far, VCGs do look like they will work fine on GLP studies, too.”

Not every study is a good match for VCGs, though.

“One thing that we’ve never said is that the virtual control groups will be appropriate for every study and that they’ll replace control groups forever,” says O’Neill. “We don’t believe that. Our goal is to have a meaningful reduction in the number of animals that are needed, but complete replacement is far on the horizon, I think.”

For instance, it is probably too risky right not to include VCG data in designing a pivotal, first-in-human enabling trial, says O’Neill. Introducing a new biomarker, a unique anatomic site for sampling, or a new vehicle that is being used in dosing treated and control animals is also probably not a good fit for VCGS right now, either. “But as time goes on, as more and more people use those new biomarkers or those new vehicles, we will have that data available and be able to use virtual controls on those studies as well,” says O’Neill. “Even in the situations that we believe are not a good fit for a VCG, we can still use the data lake to provide curated historical control data that may help increase statistical power or make sense of rare findings, improving study quality.”

The road to commercialization for virtual control groups

How extensively VCGs fit into the nonclinical space will come down to acceptance among pharmaceutical companies and regulators. Toward that end, Charles River is looking to add to its core base of large pharmaceutical partners and even smaller pharmaceutical and biotech companies willing to test the method and provide meaningful feedback. “The more input we get from our stakeholders, the better the product we’ll be able to produce in the end,” says O’Neill, adding that he is hopeful Charles River will be able to implement its first non-GLP studies using VCGs by 2026.

While the framers of VCGs, like O’Neill, believe it can have a transformative impact on toxicology testing, not the least being the dramatic decline in animal use, he acknowledged not everyone feels the same. “There is understandably a hesitancy to move on from what we have done for decades in nonclinical safety testing. And any new methodology deserves to be challenged and thoroughly vetted before it is put into place,” says O’Neill. “Our goal really is to make sure that we're not having another thalidomide disaster. That's the whole point of nonclinical safety testing. And I will be the first one to come out on a stage and say that we were wrong if it looks like that's the way it's going. But so far, all the evidence that I see suggests that the use of VCGs is a valid approach and that we'll be able to move forward with it.”