|
Activity Number:
|
202
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #310302 |
|
Title:
|
Flexible Modeling of the Hazard Function for Breast Cancer Recurrence
|
|
Author(s):
|
Vanja Dukic and James Dignam*+
|
|
Companies:
|
The University of Chicago and The University of Chicago
|
|
Address:
|
5841 S Maryland Ave, Chicago, IL, 60637,
|
|
Keywords:
|
multiresolution models ; Bayesian ; survival analysis ; hazard estimation ; prognostic factor modeling ; breast cancer
|
|
Abstract:
|
The hazard function may reveal important aspects of the failure process not apparent in the survival function. While nonparametric hazard estimates tend to be unstable, parametric models are typically too restrictive and semi-parametric approaches treat the hazard as a nuisance parameter, focusing instead on covariate effects. Here we examine the hazard of recurrence after treatment for breast cancer, with specific interest in a) the shape of the hazard and how disease characteristics influence it and b) whether there are one or more hazard changepoints. To address these questions, we extend the Bayesian multiresolution hazard (MRH) estimator (Bouman et al. 2007, 2005). The proposed hierarchical MRH model accommodates stratum heterogeneity and common covariate effects across strata. We contrast results with those from flexible parametric model extensions and nonparametric approaches.
|
