|
Activity Number:
|
95
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 7, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #306057 |
|
Title:
|
To Pair or Not To Pair? Noninferiority Test for the Difference in Correlated Binary Outcomes
|
|
Author(s):
|
Dong Li*+ and David Breiter and Shu Han and Jia Wang
|
|
Companies:
|
Guidant Corporation and Guidant Corporation and Guidant Corporation and Guidant Corporation
|
|
Address:
|
4100 Hamline Ave., N., St. Paul, MN, 55112,
|
|
Keywords:
|
non-inferiority test ; correlated binary outcomes
|
|
Abstract:
|
Noninferiority tests are used widely in the medical device industry. Two designs are based on extreme assumptions: unpaired setting assumes independency between two responses while paired setting assumes high correlation between two responses. Nam and Blackwelder's studies show the paired design has a clear advantage over the unpaired design in terms of asymptotic relative efficiency (ARE) when two responses of the pair are highly positively correlated. However, extended power analysis for the noninferiority test is not yet available to describe how the test power changes while the correlation between two responses varies. We evaluate the noninferiority test power with correlation varying between two responses. The results are compared further with the ARE between paired and unpaired settings. Extensive Monte Carlo simulations are conducted to demonstrate our findings.
|
