Mid-course sample size modification in clinical trials based on the observed treatment effect

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• Clinical trials often set sample sizes for expected treatment effects, potentially underpowering detection of smaller, yet clinically significant, effects.
• Adaptive trial designs allow mid-course modifications to increase power for smaller effects if initial results are weak.

Purpose of the Study:

• To propose adaptive group sequential designs that maintain the sufficiency principle while increasing power for smaller treatment effects.
• To reduce expected sample size in clinical trials when aiming for high power at a small treatment effect.

Main Methods:

• Discussing existing adaptive redesign methods and highlighting their limitations regarding sufficiency.
• Developing novel group sequential designs that incorporate the possibility of a small treatment effect from the outset.
• Comparing the proposed methods with L. Fisher's 'variance spending' procedure in terms of power and expected sample size.

Main Results:

• The proposed methods achieve desired power for smaller treatment effects while adhering to the sufficiency principle.
• The suggested group sequential designs can lead to reduced expected sample sizes compared to initial specifications.
• Comparisons indicate the proposed methods outperform L. Fisher's 'variance spending' procedure in certain scenarios.

Conclusions:

• Adaptive redesigns can be effectively implemented while preserving statistical sufficiency by considering all eventualities at the design stage.
• The proposed group sequential designs offer an advantageous alternative to existing methods, balancing power and sample size efficiency.
• While mid-course flexibility is appealing, substantial increases in sample size may be required to correct initial designs.