|
Activity Number:
|
111
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #303916 |
|
Title:
|
Dynamic Predicting by Landmarking as an Alternative for Multistate Modeling: An Application to Acute Lymphoid Leukemia Data
|
|
Author(s):
|
Hans C. van Houwelingen*+ and Hein Putter
|
|
Companies:
|
Leiden University Medical Center and Leiden University Medical Center
|
|
Address:
|
P.O. Box 576, Bilthoven, International, 3720 AN, The Netherlands
|
|
Keywords:
|
multi-state model ; landmarking ; dynamic prediction
|
|
Abstract:
|
We consider the problem of obtaining dynamic x-year survival prediction in clinical cohorts. As an example we look into 5-year failure free survival after bone marrow transplantation in ALL patients using data from the EBMT, the European Group for Blood and Marrow Transplantation. We compare the new landmark methodology as developed in [1] and the established multi-state modeling as described in [2] and compare the predictions coming from both approaches. As expected the two approaches give quite similar results. The great advantage of the landmark methodology is that it does not does not need complex modeling and leads to easy prediction rules. On the other hand, it does not give the insight in the biological processes as obtained from the multi-state model.
|
