Sizing a Phase IIb Trial: Using the Predictive Values from a Nonlinear Model

• Activity Number: 326
• Type: Invited
• Date/Time: Tuesday, August 6, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Jerry John Weaver*+
• Companies: Novartis Pharmaceuticals Corporation
• Keywords: bootstrap ; non-linear models ; predicted values ; phase IIb ; dose-response

Abstract

The need to design rigorous phase IIb trials with the clear objective of selecting the most appropriate doses to carry into a phase III program is a crucial step in the drug development process. Getting the dose wrong in phase III could result in late stage drug development failure or an approved drug with doses associated with post marketed surveillance concerns.

In many situations, sizing a phase IIb trial is based upon hypothesis testing rather than estimation. Statistical inferences such a trend tests are commonly used to justify sample size requirements, but the question of "what are the doses to carry into phase III?" is not adequately addressed.

An applied method is proposed for assessing the number of doses to have in a phase IIb trial, as well as sizing the dose arms such that one can reliably estimate the doses to carry into a future phase III trial. This methodology relies on obtaining a bootstrap distribution for dose given a targeted predictive value from a non-linear dose response model. Simulations will be performed to assess bias and the precision of this method when assuming very basic dose response profiles. Issues of non-convergence will also be discussed.