Power Consideration in Clinical Outcome Study with Uncertain Endpoints

• Activity Number: 487
• Type: Contributed
• Date/Time: Thursday, August 11, 2005 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Hongwei Wang*+ and Cong Chen and Steven Snapinn
• Companies: Merck & Co., Inc. and Merck & Co., Inc. and Amgen, Inc.
• Keywords: Cox regression ; Log-rank test ; Power ; Endpoint adjudication ; Poisson process

Abstract

Clinical outcome studies play a central role in providing hard evidence to draw concrete conclusions about the efficacy and safety of candidate drugs. The conventional procedure is to classify each potential endpoint as true or false by an Endpoint Adjudication Committee (EAC) and include only the confirmed endpoints in a Cox model. This strategy has potential disadvantages: the result may vary depending on the criterion chosen by the EAC; the confirmed endpoints are treated equally of being true and the uncertainty information is lost. By introducing the probability of being true for each potential endpoint, we identified the relationship between the predicted power of the trial and the threshold chosen by the EAC, which leads to an optimal criterion maximizing the power. Another weighted method (Snapinn 1998) makes use of all the potential endpoints by assigning weights based on their probabilities of being true. Under a reasonable framework, we show the weighted method is more powerful than the conventional procedure for any level of the threshold. The results are confirmed by a simulation study and applied to the planning of a cardiovascular endpoint clinical trial.