Related Experiment Videos
Early stopping to accept H(o) based on conditional power: approximations and comparisons
1Department of Biostatistics, Harvard School of Public Health, Boston, Massachusetts 02115, USA.
Biometrics
|September 18, 1997
Summary
Clinicians can stop clinical trials early if results suggest the null hypothesis (Ho) is likely. This study explores sequential testing procedures and their power approximations for improved trial design.
Area of Science:
- Clinical Trials
- Biostatistics
- Statistical Inference
Background:
- Early stopping of clinical trials is appealing when the null hypothesis (Ho) appears likely.
- Existing methods for conditional power calculation include those by Lan et al. (1982) and Pepe & Anderson (1992).
Purpose of the Study:
- To investigate a modification of the Pepe and Anderson (1992) procedure for early trial stopping.
- To formally define and study sequential tests with lower stopping boundaries.
- To improve understanding of parameter interactions and power approximation in clinical trial design.
Main Methods:
- Formal definition of sequential tests with lower stopping boundaries.
- Utilizing visual displays to illustrate parameter interplay.
- Approximating statistical power as a boundary crossing probability.
Main Results:
- Comparison of different sequential testing designs under various parameter configurations.
- Demonstration of a modified Pepe and Anderson (1992) procedure with enhanced interpretability.
- Improved understanding of conditional power calculations in early stopping rules.
Conclusions:
- The study provides a framework for formalizing and comparing early stopping procedures in clinical trials.
- Visualizations and boundary crossing approximations enhance the understanding of sequential test performance.
- The modified procedure offers a more interpretable approach to conditional power analysis for trial decision-making.