Randomization as a basis for inference in noninferiority trials.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• Noninferiority testing is typically analyzed within Neyman-Pearson or Bayesian frameworks.
• The role of randomization in clinical trials is crucial for both design and analysis.
• Existing discussions on noninferiority testing often overlook the Fisherian perspective.

Purpose of the Study:

• To present a Fisherian framework for noninferiority testing in clinical trials.
• To explore the application of permutation tests in active control noninferiority analyses.
• To deepen the understanding of the historical and philosophical underpinnings of noninferiority testing.

Main Methods:

• Utilizing a Fisherian inferential framework, which relies solely on the assumption of randomization.
• Investigating the use of permutation tests, contingent on exchangeability assumptions.
• Analyzing active control noninferiority scenarios.

Main Results:

• Demonstrates that randomization is the sole assumption required for inference in the Fisherian framework.
• Highlights the potential utility of permutation tests in active control noninferiority analyses.
• Provides insights into the historical development of noninferiority testing methodologies.

Conclusions:

• The Fisherian framework offers a distinct perspective on noninferiority testing, centered on randomization.
• Permutation tests can be valuable tools in specific noninferiority testing contexts.
• This work contributes to a more comprehensive understanding of noninferiority testing, including complex issues like intention-to-treat analysis.