|
Activity Number:
|
195
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #300811 |
|
Title:
|
Prediction in Multivariate Survival Models
|
|
Author(s):
|
M. Brent McHenry*+ and Stuart R. Lipsitz and Debajyoti Sinha and Sundar Natarajan
|
|
Companies:
|
Bristol-Myers Squibb Company and Harvard Medical School and Florida State University and VA New York Harbor Harbor Healthcare System
|
|
Address:
|
5 Research Parkway, Wallingford, CT, 06492,
|
|
Keywords:
|
Multivariate survival analysis ; full-likelihood approach ; prediction ; oncology
|
|
Abstract:
|
Multivariate survival data arise from time-to-event studies when either of two or more types of events occur for the same subject. In these studies, survival times are correlated within subject, violating the independence of survival times assumption otherwise required. A popular approach for analyzing such data is the marginal approach. We propose a full likelihood approach to estimate the conditional risk, given the history of another event. We evaluate the effect of treatment on the risk of one outcome given the history of other events and covariates. We estimate the joint distribution of the survival times, and the conditional distribution for one survival time given the other survival times. This new approach allows one to predict risk for one particular event, given the past history of other events. The method will be illustrated using an oncology example.
|
