|
Activity Number:
|
223
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistical Consulting
|
| Abstract - #304983 |
|
Title:
|
Estimating the Minimally Significant Difference for a Clinical Study
|
|
Author(s):
|
Robert A. Parker*+
|
|
Companies:
|
Harvard School of Public Health
|
|
Address:
|
Center for Biostatistics in AIDS Research, Boston, MA, 02115,
|
|
Keywords:
|
sample size estimation ; alternative hypothesis ; clinically important difference ; detectable difference
|
|
Abstract:
|
Although a number of factors are involved in calculating a sample size for a clinical study, perhaps the most critical issue is the difference to be detected, sometimes termed the "minimally significant difference" (MSD). This is the alternative hypothesis used for the study design. In practice, sample sizes often are calculated iteratively and assumptions modified---particularly the MSD---until the sample size is acceptable. I present a somewhat more objective method for determining an appropriate MSD for a study. I argue the MSD should be related to the overall potential of the study to impact clinical practice, both in terms of the potential number of future individuals impacted by the results of the study and the practical implications of such a change. Several studies in HIV/AIDS are used to illustrate the underlying ideas.
|
