|
Abstract:
|
Selecting a dose or doses in drug development for confirmatory studies is typically performed using evidence from a randomized controlled dose-finding study. Some clinical trials in autoimmune diseases present challenges to this process. First, as large numbers of treatments are approved or promising treatments are in concurrent confirmatory testing, enrollment of patients to dose-finding studies is challenging. Second, traditional assumptions on the underlying dose-response trend are uncertain in some treatments that impact human immune response. Adaptive dose-finding trials allow for more efficient use of patients in this setting, driven through quick stops of ineffective treatments and adaptive allocation of patients away from doses that are not likely to be useful given observed trial data. Additionally, Bayesian Model Averaging across potential candidate dose-response models protects against model misspecification and reduces potential bias. This presentation provides simulation results which are used to compare several design options for an adaptive dose-finding study with Bayesian Model Averaging, demonstrating a real experience of clinical trial design under constraints.
|
