Abstract:
|
In recent years, Bayesian design and analysis have generated extensive discussions in clinical trials. When using Bayesian methods, prior distributions are required. However, the use of co-data (trial-external relevant information) into prior is challenging. In this work we propose an approach to use relevant historical control data for designing trials with time-to-event endpoints. The proposed approach captures historical data for control group in a meta-analytic framework by considering the between-trial variability (Spiegelhalter et al. 2004, Neuenschwander et al. 2010, Gsteiger et al. 2013, Schmidli et al. 2014). The methodology is flexible to incorporate individual as well as aggregated co-data and differential discounting. We use a phase II Oncology trial design to illustrate the methodology along with essential practical aspects.
|