|
Activity Number:
|
177
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 4, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #301707 |
|
Title:
|
Normal Dynamic Linear Model for Dose-Finding in a Phase II Bayesian Adaptive Trial
|
|
Author(s):
|
Jason T. Connor*+ and Scott M. Berry and Donald A. Berry
|
|
Companies:
|
Berry Consultants and Berry Consultants and The University of Texas M.D. Anderson Cancer Center
|
|
Address:
|
2534 Lake Debra Drive, #108, Orlando, FL, 32835,
|
|
Keywords:
|
Adaptive trial ; NDLM ; Bayesian ; Biostatistics ; Clinical trial
|
|
Abstract:
|
We describe a normal dynamic model for dosing finding. This Gaussian process model is a flexible not necessarily monotonic model for outcomes primarily defined by one parameter (e.g., mean of a continuous response, Bernoulli probability, mean survival). Using the NDLM we estimate the probability distribution of the dose response curve and adaptively randomize to efficiently identify the dose or doses of interest (minimally effective dose, ED90, maximum utility dose, etc). The dose-response curve has no forced parametric form however there is sharing across doses governed by the drift parameter. We compare the NDLM to common parametric dose-response curves and demonstrate the flexibility and strength of the NDLM model.
|
