Related Experiment Videos
Over-ruling a group sequential boundary--a stopping rule versus a guideline
K K Gordon Lan1, John M Lachin, Oliver Bautista
1Pfizer Global Research and Development, 50 Pequot Avenue, New London, CT 06320, USA.
Statistics in Medicine
|October 21, 2003
Summary
This study explores continuing clinical trials after crossing stopping boundaries. Buying back alpha maintains type I error and power, offering a simplified approach for trial monitoring.
Area of Science:
- Biostatistics
- Clinical Trial Design
Background:
- Group sequential designs allow early trial stopping for efficacy or futility.
- Continuing trials after crossing efficacy boundaries necessitates re-evaluating statistical plans.
Purpose of the Study:
- To evaluate group sequential procedures that continue trials after efficacy boundaries are crossed.
- To assess the impact of 'buying back' spent alpha on trial properties.
- To propose simplified monitoring procedures.
Main Methods:
- Analysis of group sequential procedures with O'Brien-Fleming-like and Pocock-like spending functions.
- Evaluation of 'buying back' previously spent alpha probability.
- Comparison with fixed-sample size critical values for future looks.
Main Results:
- O'Brien-Fleming-like spending functions show negligible impact on type I error and power.
- Pocock-like bounds result in a small power loss when buying back alpha.
- Using fixed Z critical values (e.g., 1.96) preserves type I error with minimal power loss.
Conclusions:
- Continuing trials after crossing efficacy boundaries is feasible with 'alpha buy-back' strategies.
- Simplified procedures using fixed critical values maintain trial integrity.
- This approach provides a robust stopping boundary rather than a mere guideline.