|
Activity Number:
|
246
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #308287 |
|
Title:
|
Conditional Power for Stratified Logrank Tests
|
|
Author(s):
|
Chung-Kuei Chang*+ and Lothar Tremmel
|
|
Companies:
|
Cephalon, Inc. and Cephaloh, Inc.
|
|
Address:
|
18 Moores Rd, Malvern, PA, 19355,
|
|
Keywords:
|
conditional power ; logrank test ; stratification ; survival analysis ; interim analysis
|
|
Abstract:
|
To make a go/no-go decision at interim analyses in confirmatory clinical trials, it is important to predict the chance of success at the final analysis based on the interim data and various assumptions in the future data. Lan and Wittes (1988) proposed the calculation of the conditional power based on the B-value theory, in which the expectation and variance of the conditional Z statistic at the final analysis were derived. However, their theory may not apply to stratified analyses. The issue is how to weight each stratum appropriately in order to estimate the variance of the pooled Z statistic. In the analysis of time-to-event data using the stratified logrank test, we provide a precise derivation to calculate the conditional power based on the decomposition of the logrank statistic proposed by Tsiatis (1981).
|
