Exact unconditional tests for testing non-inferiority in matched-pairs design

Area of Science:

• Biostatistics
• Clinical Trials
• Statistical Methodology

Background:

• Non-inferiority testing is crucial for evaluating new treatments against existing ones.
• Matched-pairs designs are common in clinical research but present unique statistical challenges.
• Existing asymptotic tests for non-inferiority may lack accuracy, especially with smaller sample sizes.

Purpose of the Study:

• To develop and evaluate exact unconditional tests for non-inferiority in 2x2 matched-pairs samples.
• To compare the performance (size and power) of proposed exact tests against existing asymptotic methods.
• To provide accurate statistical tools for non-inferiority assessment in clinical trials.

Main Methods:

• Development of two exact unconditional tests utilizing standard and confidence interval p-values.
• Reduction of nuisance parameter space dimensionality from two to one through monotonicity.
• Exact size and power calculations for proposed and existing tests.

Main Results:

• The proposed exact unconditional tests demonstrate accurate size properties.
• The exact test based on the confidence interval p-value is more powerful than the alternative exact test.
• The existing asymptotic test exhibits inaccurate size, exceeding the nominal alpha level.

Conclusions:

• Exact unconditional tests are recommended for accurate non-inferiority testing in 2x2 matched-pairs designs.
• The confidence interval p-value based exact test is the preferred method due to its higher power.
• Caution is advised when using asymptotic tests for non-inferiority, particularly with small to moderate sample sizes.