Carl, Lilly ran a phase II/III Bayesian trial for its diabetes drug, Trulicity. Was a Bayesian phase II trial design, which picked 1 or 2 doses based on a Bayesian algorithm and continued to phase III with the selected doses. The seamless switch also started other phase III trials. This is a nice example because it has FDA approval and Lilly states it saved them 12-18 months, with $337M last quarter; exceeding projections of $1.2B/year, and of course the year plus savings is not small. A bigger story is that the phase II used 7 doses and response adaptive randomization, rather than the preplanned 3 doses. They may have ended with a dose they would never have explored. Two papers attached. Ferring is running a sepsis trial with a Bayesian adaptive phase II which could drop arms and move to phase III. Trial still ongoing.
ClinicalTrials.gov Identifier: NCT02508649
https://clinicaltrials.gov/ct2/show/NCT02508649?term=Sepsis-act&cond=Sepsis&rank=1
A paper about this paper should be coming out soon.
As agreed, I looked briefly at publicly available info on Bayesian designs accepted by FDA. There is a lot in device space, including recent RCT in Lancet, but limited number of examples we could site for CDER/CBER: