A Dunnett-Bonferroni-based parallel gatekeeping procedure for dose-response clinical trials with multiple endpoints

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methods

Background:

• Existing Dunnett-based parallel gatekeeping procedures in dose-response trials with multiple endpoints rely on assumptions difficult to justify.
• These assumptions include approximating joint distributions with multivariate-t and accurately estimating correlations, which can be challenged by regulatory agencies.

Purpose of the Study:

• To propose a novel Dunnett-Bonferroni-based parallel gatekeeping procedure for dose-response clinical trials with multiple endpoints.
• To relax restrictive assumptions of existing methods while maintaining high statistical power and ease of implementation.

Main Methods:

• The proposed method adapts the Dunnett-based parallel gatekeeping strategy by incorporating the Bonferroni inequality.
• Type I error rate is split among families, avoiding the need for multivariate-t distribution approximations or precise correlation estimations.

Main Results:

• The Dunnett-Bonferroni-based procedure avoids assumptions that may face regulatory scrutiny.
• While potentially slightly less powerful than the Dunnett-based method, the power loss is minimal.
• The proposed method is generally easier to implement than the traditional Dunnett-based procedure.

Conclusions:

• The Dunnett-Bonferroni-based parallel gatekeeping procedure offers a robust and practical alternative for clinical trials with multiple endpoints.
• It provides a statistically sound approach that is more amenable to regulatory review and simpler for researchers to apply.