Update Bayesian Predictive Power after a blinded interim analysis


Assume an interim analysis for efficacy is performed, using a pre-specified interim boundary, e.g. that the hazard ratio must be below a pre-set threshold. However, the statistician remains blinded to the result in the sense that she only receives the information that the interim has been passed, but will not know the precise effect estimate. We discuss methodology how to quantify that knowledge and use it to update Bayesian predictive power.

Roche Global Biostatistics Conference (contributed talk)
Brighton, United Kingdom