|
Activity Number:
|
233
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #304562 |
|
Title:
|
Biased Results from Single-Arm Phase II Cancer Clinical Trials: Simulations Based on a Large Clinical Trial
|
|
Author(s):
|
Hui Tang*+ and Nathan R. Foster and Richard M. Goldberg and Daniel Sargent
|
|
Companies:
|
Mayo Clinic and Mayo Clinic and The University of North Carolina at Chapel Hill and Mayo Clinic
|
|
Address:
|
200 First St SW, Rochester, MN, 55905,
|
|
Keywords:
|
Phase II clinical trial ; Randomization
|
|
Abstract:
|
The historical standard for phase II trials in oncology has been a single arm trial, with results compared to historical control. We illustrate sources of bias and the resultant impact on trial conclusions using both simulations and real data from large colorectal cancer studies. We demonstrate that even in the presence of a modest error (5%) in the assumed historical control success rate, the false positive rate in single-arm designs is inflated two to three fold, while the randomized two-arm design retains the desired error rates. Increasing the sample size in the single arm design only further inflates the Type I error rate. We also demonstrate from actual trials that selection of patients from different treating locations greatly influences the error rates, and using a null based on historical studies rather than a prospective control profoundly impacts trial conclusions.
|
