Changing Course Midstream or Giving Up Entirely: Making Decisions on the Basis of Ongoing Data from Clinical Trials

• Activity Number: 636
• Type: Topic Contributed
• Date/Time: Thursday, August 13, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Janet Wittes*
• Companies: Statistics Collaborative
• Keywords: Clinical trials ; conditional power ; data monitoring ; adative designs

Abstract

Many trial protocols include specify methods for evaluating data with the view toward stopping for futility. Other trials specify methods for adaptive modifications in response to interim analysis. When the protocol is written with statistical care, these potential changes are structured in ways that preserve the trial's statistical operating characteristics. Implementation of these protocols may be more complicated than the designers of the study anticipated. Data may emerge that differ from any scenario envisioned by the protocol's authors. The standard of care may change while the trial is ongoing. The data monitoring committee (DMC) charged with making recommendations about these designed changes may be uncomfortable proceeding as specified. Methods based on conditional may produce power that appears intuitively much lower, or higher, than what the data appear to imply. This talk proposes a 2-dimensional grid describing the types of changes a DMC might recommend and the degree of regulatory and scientific risk these potential changes incur. It addresses, with actual examples, the types of conjectures masked investigators are likely to make in response to a DMC's statements.