A comparison of two methods for adaptive interim analyses in clinical trials

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methods

Background:

• Two-stage adaptive trial designs offer flexibility in clinical research.
• Early termination is possible if interim analysis shows no treatment effect, preserving resources.
• Observed treatment effects can inform subsequent trial stages while controlling Type I error.

Purpose of the Study:

• To compare two proposed methods for two-stage clinical trial procedures.
• To evaluate their statistical performance, including Type I error rate and power.
• To assess their practical applicability and provide guidance for use.

Main Methods:

• Analysis of two-stage procedures proposed by Bauer and Köhne (1994) and Proschan and Hunsberger (1995).
• Derivation of the exact Type I error rate for the Proschan and Hunsberger method.
• Comparison of decision rules, power, and expected sample size between the two methods.

Main Results:

• Both methods allow early stopping for no treatment effect and protect the Type I error rate.
• Decision rules are similar, with negligible differences in power and sample size.
• Significant differences exist in design tools and practical applicability.

Conclusions:

• The Bauer and Köhne, and Proschan and Hunsberger methods offer comparable statistical efficiency for two-stage trials.
• Recommendations are provided based on the practical performance and design considerations of each procedure.
• Choosing between methods depends on specific design needs and practical implementation factors.