|
Activity Number:
|
64
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 6, 2006 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #306927 |
|
Title:
|
Allowing for Etiologic Heterogeneity by Disease Subtype Increases the Power of Tests for Genetic Association
|
|
Author(s):
|
Peter Kraft*+ and Sholom Wacholder and Nilanjan Chatterjee
|
|
Companies:
|
Harvard University and National Cancer Institute and National Cancer Institute
|
|
Address:
|
Departments of Epidemiology and Biostatistics, Boston, MA, 02115,
|
|
Keywords:
|
genetic epidemiology ; categorical outcome ; polytomous regression
|
|
Abstract:
|
Many complex diseases can be broken into several pathological or clinical subtypes. A putative disease susceptibility gene may influence risk of one, several, or all of these subtypes. We present two simple procedures based on polytomous logistic regression to test the global null hypothesis that a polymorphism is not associated with variation in disease outcome. The first compares the null model to a saturated alternative; the second uses a two-stage model to reduce the number of free parameters in the alternative. When there is no subtype-specific genetic effect, these procedures are only slightly less powerful than standard dichotomous logistic regression procedures; when there is a subgroup-specific effect, these procedures can be dramatically more powerful. We illustrate these procedures with an application to a case-control study of breast cancer, classified by ER and PR status.
|
