|
Activity Number:
|
528
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Thursday, August 2, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #308920 |
|
Title:
|
Time Dependence of a Gene-Expression Prognosticator from Modeling Breast Cancer Survival Data
|
|
Author(s):
|
Kit Fun Lau*+ and John Sninsky and Trevor Hastie
|
|
Companies:
|
Celera and Celera and Stanford University
|
|
Address:
|
5319 Matthew Terrace, Fremont, CA, 94502,
|
|
Keywords:
|
time dependent prognostic ; breast cancer ; gene expression signature ; time dependent AUC ; metastasis score
|
|
Abstract:
|
An RT-PCR based gene-expression prognostic signature has been derived to predict distant metastasis for breast cancer. Supervised principal components and LASSO methods were used to carry out the gene selection and form a metastasis score (MS) which is a multivariate index predictor. The signature has subsequently been validated in two independent datasets. Moreover, the prognostic values of the molecular signature were shown to be most accurate for the first five years but to decrease over time. The time dependence was manifested by 1) a decrease in hazard ratios between MS high-risk and low-risk groups over time, 2) trend toward statistical significance for the time-dependence of the regression coefficients of MS in a Cox model and 3) a decrease in time-dependent AUC. Temporal evolution of the molecular prognosticator will shed light on its usefulness for different time horizons.
|
