A bootstrap-based test for establishing noninferiority in clinical trials

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmacoeconomics

Background:

• Noninferiority trials compare a new treatment to an active control.
• Establishing noninferiority often relies on confidence interval methods.
• Existing methods may not preserve the nominal significance level.

Purpose of the Study:

• To evaluate a novel bootstrapping method for noninferiority testing.
• To improve the accuracy of significance level approximation in noninferiority trials.
• To enhance the statistical power of noninferiority tests for continuous outcomes.

Main Methods:

• A randomized, active-control clinical trial design was considered.
• Noninferiority margin was defined based on preserving a fraction of the active control effect.
• Bootstrapping was employed to derive accurate confidence interval limits.

Main Results:

• The proposed bootstrapping method better approximates the nominal significance level.
• This approach leads to improved statistical power for noninferiority testing.
• Traditional confidence interval-based tests may not maintain the intended significance level.

Conclusions:

• Bootstrapping offers a more accurate and powerful approach for noninferiority testing in clinical trials.
• This method ensures better control over Type I error rates.
• It provides a more reliable assessment of treatment noninferiority for continuous variables.