The first goal of this project is to develop Bayesian statistical methods for using historical information in analyzing data from clinical trials. This has most relevance in those settings in which additional randomized data are difficult to obtain for ethical or other reasons. In the context of two actual medical settings the plan is to: (1) develop the model suggested in the proposal,(2) elicit necessary probability distributions,(3) write computer software to carry out the necessary calculations,(4) compare with existing metaanalysis procedures. The second goal is to carry out a decision making process in the context of a prophylactic trial. This will entail eliciting probability distributions and loss functions from the investigators and others. These will be used to evaluate the various possible actions, such as in deciding whether or not to stop a clinical trial.
| Lewis, Roger J; Lipsky, Ari M; Berry, Donald A (2007) Bayesian decision-theoretic group sequential clinical trial design based on a quadratic loss function: a frequentist evaluation. Clin Trials 4:5-14 |
