By Deborah Borfitz
March 17, 2008 | Lack of regulatory guidance may be the main, if not the only, barrier to widespread adoption of adaptively designed clinical trials across study phases.
The statistical and technological know-how necessary to conduct adaptive clinical trials (ACTs) already exists. PC-compatible, open-source software known as WinBUGS accommodates the Bayesian statistical method with which ACTs have become more or less synonymous, says Jay Herson, PhD, senior associate in biostatistics at Johns Hopkins University. Simulation technology using programming language R also exists so companies doing ACTs can adequately plan for a multiplicity of possible sample sizes, outcomes, and resource consumption scenarios.
Regulatory authorities have only to give blessing to the use of specific ACTs through guidance, says Herson. For the foreseeable future, there won’t be a shortage of sufficiently skilled Bayesian loyalists to handle the complex calculations associated with ACTs.
The big unknown is precisely how many ACTs have actually been done or are now under way, says Herson. “All the big pharmaceutical companies push for adaptive designs. PhRMA [Pharmaceutical Research and Manufacturers of America]…pretends the ‘time is here’ in hopes of intimidating the FDA politically. The strategy is not working and appropriately so.”
Based on an initial survey last year of 13 mid- and large-size pharmaceutical companies, PhRMA’s Adaptive Design Working Group came up with 37 examples of ACTs, says Judith Quinlan, co-chair of the group’s case study work stream. Of those, three were Phase I, one combined Phase I and II, fifteen were Phase II, two combined Phase IIA and IIB, nine combined Phase II and III, four were Phase III, and three were Phase IV.
“All but one focused adaption on dose,” says Quinlan. The exception adjusts on population. Nine of the 37 used Bayesian statistics, including seven of the fifteen Phase II trials. The combined Phase II and III adaptive designs were a mix of “operationally seamless” trials and those combining information from two phases. The survey also found that FDA interactions only occurred with the Phase II examples if the information was expected to be used for a regulatory submission.
The admittedly biased sample underrepresented early phase trials, which respondents saw as less “interesting,” and late-phase confirmatory trials, due to confidentiality issues, says Quinlan. “I don’t think we’ll have clean metrics for some time.”
Additional case studies are now being collected from clinical research organizations working on behalf of smaller companies, says Quinlan. The new case studies include about 20 trials (half for devices) contributed by Don Berry, chairman of the department of biostatistics and applied mathematics at the University of Texas M.D. Anderson Cancer Center, and another 10 or so contributed by biostatistical consultancy and software company Cytel.
Among them will be a large Phase III “population enrichment” trial involving platelet inhibition for acute coronary syndrome being done by Cytel on behalf of The Medicines Company, a small pharmaceutical company based in Parsippany, NJ. Subjects are sub-grouped according to their medical condition, and the drug gets randomized only to those groups it is found to benefit, says Cytel President Cyrus Mehta. The trial can adapt “up to 15,000” patients.
------------------------
This story first appeared in eCliniqua,one of Bio-IT World’s free e-newsletters. Subscribe here.