|
Activity Number:
|
343
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 11, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #300876 |
|
Title:
|
Estimation of Probabilities of Adverse Events under Informative Dropouts
|
|
Author(s):
|
Masako Nishikawa*+ and Toshiro Tango and Makiko Ogawa
|
|
Companies:
|
National Institute of Public Health and National Institute of Public Health, Japan and Aventis Pharma Ltd
|
|
Address:
|
3-6 Minami 2 chome , Wako , International, 351-0197 , Japan
|
|
Keywords:
|
competing risk ; time-to-event data ; cumulative incidence function ; recurrent event ; nonparamtric estimation
|
|
Abstract:
|
In long-term treatment or in treatment with frequent severe adverse events (AE), such as those for oncology, it is important to know occurrence rate of AE across time and its severity. Conventional method is to calculate proportion of patients with AE to all patients enrolled neglecting dropout or to estimate occurrence rate of AE by Kaplan-Meier estimator assuming that dropout is noninformative. However, this assumption cannot be always true and not be validated by data. Furthermore assumed population is only hypothetical. We propose a method applying competing risk analysis by defining events of dropout prior AE and AE prior dropout. We focus on an AE one by one. We distinguish obvious non-informative censoring from other censorings that may not be noninformative. Therefore, our approach does not need independent assumption for dropout. The cumulative incidence function (CIF) of the AE by severity can be obtained by viewing categorized severities as competing risk. We estimate CIF of sequential occurrence of the same AE by forming a restricted risk sets. Some useful examples of graphical presentation are shown.
|
