An Advanced Design Framework For Clinical Development In The Era Of Cloud-Computing

For over a decade, advanced trial design techniques have promised efficient trials with accelerated timelines, reflecting the ability to quantify uncertainty and de-risk trials using adaptive tools. Despite the emergence of these complex innovative designs, the success of Phase 3 trials has continued to hover at 33% while the average time to market remains about 6 years.

While the number of trial designs and types have gently expanded, finding the optimal trial design for a specific context remains an elusive goal. Typically, a sponsor’s R&D team will identify five or six designs that are sent to a statistician. Once the statistician designs these trials, the most promising designs are optimized. The entire process can typically take several weeks.

Given that there exist tens of thousands of potential designs, beginning with half a dozen potential trials and then making incremental design changes is unlikely to locate the ideal trial for a specific situation. Rather, trial sponsors appear to be engaged in what some social psychologists call satisficing. They identify five or six designs that satisfy some basic requirements, and with minor modifications optimize across this constrained subset. Finding the optimal trial design requires optimizing across every potential trial.