- 作者：George F. Borm ; Gert J. van der Wilt ; Jan A.M. Kremer ; Gerhard A. Zielhuis
- 关键词：Power ; Sample size ; Statistical design ; Clinical trial ; Confidence interval ; Composite endpoints
- 刊名：Journal of Clinical Epidemiology
- 出版年：2007
- 期刊代码：148_08954356
- 类别：med
- 出版时间：April 2007
- 卷：60
- 期：4
- 页码：375-381
- 文件大小：213 K
A clinical trial may have multiple objectives. Sometimes the results for several parameters may need to be significant or meet certain other criteria. In such cases, it is important to evaluate the probability that all these objectives will be met, rather than the probability that each will be met. The purpose of this article is to introduce a definition of power that is tailored to handle this situation and that is helpful for the design of such trials.
Study Design and Setting
We introduce a generalized concept of power. It can handle complex situations, for example, in which there is a logical combination of partial objectives. These may be formulated not only in terms of statistical tests and of confidence intervals, but also in nonstatistical terms, such as “selecting the optimal by dose.”
Results
The power of a trial was calculated for various objectives and combinations of objectives.
Conclusion
The generalized concept of power may lead to power calculations that closely match the objectives of the trial and contribute to choosing more efficient endpoints and designs.