By Deborah Borfitz

July 16, 2019 | Eliminating a study's control arm by substituting pre-existing real-world data (RWD) is an idea that has captured the imagination of industry sponsors looking to gain efficiencies and cost savings—and reduce participants' fears that they will be end up getting a placebo. For now, the closest most trials ever get to having a so-called "synthetic control" is limited largely to historical trial data (HTD) that supplements rather than replaces the placebo group, according to Dalvir Gill, PhD, CEO of TransCelerate BioPharma.

TransCelerate enables sharing of HTD through its Placebo Standard of Care (PSoC) database, says Gill, which is generating value "unrivaled by any other project we're doing." PSoC houses control arm datasets from close to 140 trials spanning 21 disease areas, collectively representing over 85,000 patients. Sixteen of the 20 member companies are currently contributing data, with more expected to do so by the end of 2019.

The utility and value of the database is "unequivocal," says Gill. Companies are tapping the database primarily to more accurately power the control arm of studies so they can enroll fewer participants. One company, he notes, was able to increase the power of a trial's control arm from 90% to 95% while reducing the sample size by 15%.

The statistical review division of the Food and Drug Administration (FDA) has indicated control arm supplementation is, as Gill says, "an idea whose time has come." But the data must be strong and traceable as well as "apples to apples" with enrolled patients in the control arm, he adds. Supplementation is also a necessary first step to building the comfort level of sponsors and regulators with true synthetic controls.

Importantly, HTD being compiled by PSoC are neither aggregated nor cherry-picked—the data come from similar sources, curated in known ways and are consumable down to the patient level because all incoming data get converted to the same CDISC-compliant format, says Gill.

The growing database of HTD is being used by statisticians and clinicians for multiple other purposes beyond control arm supplementation, notes Gill. These range from precision powering, inclusion and exclusion criteria optimization, and improving diversity in clinical trials to safety signal interpretation, biomarker development, and disease modeling and management. One company was able to use data to counter misinformation about the toxicity of one of its marketed drugs.

The deidentified HTD will be soon be transferred from a database maintained by TransCelerate to a cloud-based platform called DataCelerate that will also house preclinical toxicology data and, ultimately, other types of currently siloed molecular, drug and therapeutic data, says Gill. "We want to be able to connect the data longitudinally and get it in the hands of clinicians and scientists… all in one place and all on one platform."

Mental Shift

Currently, adding HTD to the platform is a blinded request-and-response procedure between member companies that is trial-specific, explains Jules Desmond, PhD, strategic development director, Global Development for Amgen, who leads PSoC. Amgen alone has to date donated about 2,000 subjects' worth of data.

Company representatives in the TransCelerate workstream foresee a future where control arm data is routinely uploaded to the database and available for immediate use, shortly after studies are completed and published—"the more the merrier," says Desmond. But the feeling of companies at present is that they own the data they collect, so TransCelerate has them sign a data use agreement that lays out the ground rules for acceptable use of data contained in the database.

"We have the foundations of a pretty robust database already, but companies aren't required to provide data and they don't have to provide a reason why they don't want to share," says Desmond. "Despite that, we've seen good behavior and a willingness to share."

Companies may have legitimate reasons not to share control arm data, Desmond says, such as when the drug under study is still experimental. But HTD downloaded to the database are free and accessible to all member companies as needed to improve their clinical development programs, which is "a big mental shift for companies" that should become easier with time. Sharing of control data can also be resource-intensive for companies, Desmond adds, so they first want to fully assess the value of the practice.

For potential database users, the "devil is in the details," Desmond says. "It's not enough to just have a study in there—you need a study that matches your precise population."

The HTD contributed by TransCelerate companies have largely been in therapeutic areas outside oncology where no other data-sharing options exist, says Desmond, although that is starting to change. This could also move the group closer to the possibility of synthetic control arms, which have been utilized almost exclusively in single-arm, early-phase oncology studies.

Add or Replace

It may be unrealistic to expect companies to use HTD to completely replace control arms in studies that will be used for registration, says Desmond. But Amgen has used RWD to construct a comparative cohort, thereby accelerating the approval of Blincyto (blinatumomab) for the treatment of a rare form of leukemia. Previous pediatric clinical trials have also utilized HTD from adults to supplement the control arm and an upcoming pediatric study on another molecule will use HTD from a previous study involving children—methodologies that regulatory agencies have grown "much more amenable to."

Historical controls are a consideration for any study, when appropriate, and Amgen has teams of experts exploring all the potential data sources, says Desmond. Outside of trial data, electronic health records, administrative claims data, disease registries, and patient-generated data from fitness trackers and home medical equipment are all candidates.

Amgen has been looking at data sources outside TransCelerate to create synthetic controls in stage I oncology studies, although "we haven't found a comparably matched subject cohort to do that yet," Desmond says. The company also hopes to supplement the control arm in two different phase II studies "to spare subjects from being exposed to a placebo when they don't need to be." The indications have completed large phase III studies, and the TransCelerate database could provide access to some of the recently completed trials that include populations resembling those the company wants to enroll.

If enrolled subjects look like historical subjects, Amgen could theoretically enroll 50 less new human subjects by supplementing the control arm with perhaps a few hundred former ones, Desmond says. Enrollees would have a greater chance of getting on the active arm, plus the study would complete more quickly.

Other than imparting a different randomization ratio, control arm supplementation doesn't necessarily alter a study's basic design, says Desmond. But a company could have blinded statisticians do interim analysis on the historical dataset as the trial progresses to confirm they are highly matched to recruited subjects. If they're behaving a bit differently—perhaps because the standard of care has shifted over time—they could use Bayesian methodologies to change the weight of the historical data or use less or none of the HTD and increase subject enrollment.

Stance of Regulators

Over the past 18 months, PSoC representatives have had numerous face-to-face meetings with health agencies—including the FDA, European Medicines Agency and the UK's Medicines & Healthcare products Regulatory Agency—around the issue of control arm supplementation in registrational studies, Desmond says. For earlier phase studies, "the risk is with the company… regulators don't care so much when you're not trying to make claims off of [a study]."

For registration studies, "the biggest needle mover," regulators are enthusiastic but moving more cautiously, he continues. Meetings with the FDA culminated with a cross-industry, multi-stakeholder workshop in May 2018 in Bethesda, Maryland. One of the agency's biggest concerns is "how to understand and minimize the bias introduced when you select historical data." Another, he adds, is how it can be certain that de-identified data reflect the original data from real clinical trial subjects and haven't been altered in any way.

No guidance has yet to emerge on the use of HTD for control arm supplementation, but the clear message from the FDA is that it's open to discussing the topic and will "take each proposal on its own merits," says Desmond.