Adaptive Study Design May Become Survival Tactic

By Deborah Borfitz 

December 3, 2012 | In the not-so-distant future, the pharmaceutical industry will need to move away from serendipity and its “blockbuster” drug development approach to become a service-based business specializing in patient-specific therapeutics delivering desired outcomes at the desired price. Clinical trials, of necessity, will be re-engineered to reflect real-world circumstances and turn adaptive study programs into a competitive advantage. Practicing clinicians will be dissuaded from using high-cost medicines and more actively participate in modeling and simulation exercises to help get better products to patients sooner.

Those were a few of the takeaways from an October conference, “Transforming Drug Development: Delivering Innovation and Efficiency Through Adaptive Clinical Trials,” held in Bethesda, Maryland. Top champions of the adaptive approach served as speakers, including Don and Scott Berry of Berry Consultants; Tom Parke from Tessella; Vlad Dragalin and Andy Grieve from conference host Aptiv Solutions; Michael Krams from Johnson & Johnson; and Brenda Gaydos from Eli Lilly. 

The big picture is this: the Affordable Care Act (ACA) will control the market and tie drug pricing to outcomes, resulting in $155 billion in lost revenues over the next decade, says Steve Arlington, a PricewaterhouseCoopers (PwC) global advisory leader. By 2020, PwC predicts a new regulatory environment will be needed to deal with a deluge of data. “Your drug can’t just work better than a competitor’s drug in a clinical trial, because [regulators] will have databases…with tens of thousands of pieces of data telling them outcomes on a population-size basis. They’ll know which drugs act better, probably by genotype by then.” The information will eventually become “more valuable” than the products themselves.

Real-world outcomes data is currently of limited use to pharmaceutical companies in their conversations with patients, physicians, and purchasers, says Scott Gottlieb, research fellow with the American Enterprise Institute and former Food and Drug Administration (FDA) deputy commissioner. Companies won’t risk an FDA warning letter by saying anything about a drug that isn’t in the approved labeling. The “only way” to spur such dialogue is via adaptive clinical trials (ACTs) approximating real-world circumstances and incorporating tests of significant questions.

Clinical trial requirements will in any case continue to grow, due to efficacy rather than safety concerns, Gottlieb says. The push is coming from medical reviewers driving FDA decision-making who want “objective evidence that demonstrates benefit beyond any doubt.” It’s a reflection of pressure from consumer groups and payers as well as “a growing view within the FDA that doctors can’t always be trusted to make smart prescribing decisions.”

Many Unrealized GainsAdaptive design experts at the conference were unified in their belief that predictive modeling and simulation activities improve trials—including the fixed design ones. “We found that most of the time teams go through this process, they end up with something different in their plan than they started with, whether it’s number of doses, sample size, or type of design,” says Gaydos, head of clinical trial optimization at Lilly. A “living…advanced analytic strategy document” associated with every compound ensures the necessary information for simulation and modeling is at hand when needed. Tools align with various models—for virtual patient responses, trial execution, design, final data analysis, performance metrics, and dropout rates—to help with simulation.

Simulation “very quickly” took one study team at Lilly from the standard three doses to six even before it settled on a response-adaptive trial design with Bayesian analysis. An operationally seamless phase II/III design also helped Lilly save one year over the standard clinical plan. Lilly has now started simulating entire clinical plans. Trials are simulated in sequence using information gleaned from preceding trials. “Too much emphasis has [heretofore] been put on the go/no go decision criteria” versus moving forward in the right direction, she says.

But Lilly is clearly an outlier. Based on postings to clinicaltrials.gov, the medium number of dose ratios in clinical studies is still three—“not enough to learn about dose response,” says Grieve, senior vice president of clinical trial methodology at Aptiv Solutions, a clinical research organization specializing in ACTs. Individual companies simply don’t do phase IIb trials often enough to invest in the infrastructure to support more complex study designs. The switch to adaptive dose-response trials is thus dependent on external suppliers, such as Aptiv Solutions, for the requisite substructure.

Phase II and IIb trials are the “sweet spot for adaptive designs,” says Grieve. But seamless phase IIb/III studies also have noteworthy benefits for sponsors beyond the obvious time savings. “Usually, you can use data from the phase II part in the final analysis, which is not the case when you run separate trials.”

Adoption ChallengesEven the best laid plans can sometimes be upended by data monitoring committees (DMCs). “It’s important that [DMC members] know they don’t design the trial or redesign the trial and they don’t change the sample size except as indicated prospectively in the protocol,” says Don Berry, senior statistical scientist at Berry Consultants. “I’ve seen them out of control.”

Among other ACT implementation challenges are the tediousness of arriving at the most appropriate adaptive design and quickly acting on incoming data, says Dragalin, senior vice president of Aptiv Solutions’ Innovation Center. It is helpful to run data in parallel across functional units, he adds, noting that his company offers a single multi-user platform for electronic data capture, randomization, and drug supply that is particularly well suited to the customized nature of ACTs.

Overall, sponsors need to start working in tighter partnership with academic medical centers, research institutes, payers, providers, regulators, patients—and each other—to better understand diseases and get drugs into man earlier, says Arlington. A decade from now, data will be generated not by pharmaceutical companies but across an integrated health care system with common standards, semantic technology, and predictive simulation. “You can’t do it on your own…The partnership succeeds [or fails] together.”

Patients will “start insisting” on proof-of-outcome data by 2020, says Arlington. They are also emerging as the most likely stakeholder group to broadly campaign for ACTs. Potential champions have left the FDA, says Gottlieb, and many clinical leaders with proximity to the agency are unwilling to speak up on the issue because it might be perceived that they “do the bidding” of drug makers. Krams readily agrees, citing the need for “a more objective [societal] platform…to place the solution on.”

Incorporating adaptive trials into drug development programs can both help lower clinical development costs and yield more information, says Gottlieb. But the FDA’s bifurcated review process makes it vital to offer “comfort” to medical reviewers, who typically are not well versed with statistics, about novel aspects of new drug applications.

At the sponsor level, clinical teams—even more than statisticians—tend to be enthusiastic about ACTs because it reflects the way they think, says Berry. Investigators’ concerns about patient management and the potential for bias will be overcome largely by involving them in simulation and modeling exercises, says Parke, head of clinical trial solutions at Tessella.