|
Abstract:
|
In dose-finding studies, optimal designs for estimating target doses depend on unknown dose-response model parameters. Although vague prior information might be available (e.g., from earlier studies), the model parameters cannot be determined with high precision prior to the start of a trial. An appealing idea is to estimate the model parameters at an interim analysis and use them to calculate an optimal design for the next cohort of patients. A challenge with this approach is that dose-response data are typically highly variable, resulting in unreliable parameter estimates. We use a Bayesian shrinkage approach, which utilizes vague prior information, leading to a more reliable estimation of model parameters. Motivated by design considerations for a real clinical trial we perform an extensive simulation study to evaluate the benefits of a response adaptive dose-finding approach.
|
