|
Activity Number:
|
177
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 4, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #301364 |
|
Title:
|
Comparing a Bayesian Approach with the Frequentist T-Statistic Method in Adaptive Dose-Finding Trials
|
|
Author(s):
|
Nitin R. Patel*+ and James Bolognese and Inna Perevozskaya and Robert B. Smith
|
|
Companies:
|
Cytel Inc. and Merck Research Laboratories and Merck Research Laboratories and Cytel Inc.
|
|
Address:
|
675 Massachusetts Avenue, Cambridge, MA, o2139,
|
|
Keywords:
|
adaptive ; dose finding ; Bayesian ; t-statistic method ; frequentist ; simulation
|
|
Abstract:
|
There are several adaptive designs for dose-finding trials that modify randomization ratios in cohorts based on accumulating information of responses to improve on standard fixed ratio designs. We will describe results from simulation experiments that compare two adaptive methods: a Bayesian method that uses a variance reduction criterion to adaptively allocate doses developed by Scott Berry, and the t-statistic method of Ivanova, Bolognese, and Perevozskaya. The methods will be compared for data generated from a four-parameter logistic mean response curve for both continuous and binary endpoints. Effectiveness along dimensions such as power, bias, dose-response estimation and probability of finding the target dose will be examined. Two definitions of the target dose will be investigated: the dose that gives a specified magnitude of response, or a specified difference from placebo.
|
