Activity Number:
|
294
|
Type:
|
Contributed
|
Date/Time:
|
Tuesday, July 31, 2007 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Bayesian Statistical Science
|
Abstract - #308396 |
Title:
|
Bayesian Screening for Pharmacogenetic Effects in Clinical Trials
|
Author(s):
|
Mengye Guo*+ and Daniel F. Heitjan
|
Companies:
|
University of Pennsylvania and University of Pennsylvania
|
Address:
|
503 Blockley Hall 423 Guardian Drive, Philadelphia, 19104,
|
Keywords:
|
Bayes factor ; importance sampling ; bupropion ; pharmacogenomics ; single-nucleotide polymorphism ; Laplace approximation
|
Abstract:
|
Pharmacogenetics is concerned with the detection of genetic markers that modify treatment effects. Thus, statistical methods for pharmacogenetics aim to assess the significance of treatment-by-marker interactions. When the number of potential markers is large we encounter the problem of multiplicity. We propose a Bayesian hypothesis testing method (Berger, 1985) to screen a large pool of markers for statistically significant interactions; the method evaluates evidence using Bayes factors. We carry out the computations using both importance sampling and analytical approximation. The Bayesian method explicitly incorporates prior information. Moreover, it is less conservative than a frequentist test with a Bonferroni-type correction. We apply our method to a randomized trial of pharmacotherapy for smoking cessation, in which 84 SNPs were evaluated as potential pharmacogenetic markers.
|