Sample Size and Power Calculations With Win Measures Based on Hierarchical Endpoints.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methods

Background:

• Win measures (win ratio, win odds, net benefit, DOOR) are increasingly used for hierarchical endpoints in clinical studies.
• Current sample size and power calculations often rely on cumbersome simulations or hard-to-elicit parameters.
• Investigator-specified clinically significant win measures and tie probabilities are challenging to determine from existing literature or preliminary data.

Purpose of the Study:

• To develop novel formulas for sample size and power calculations for four common win measures.
• To provide methods for computing overall win measures and tie probabilities from readily available marginal specifications.
• To enable more accessible and justifiable sample size and power estimations for hierarchical endpoints.

Main Methods:

• Derivation of sample size and power calculation formulas for four win measures.
• Development of formulas to translate marginal win measures and tie probabilities into overall measures.
• Extensive simulation studies to validate the accuracy of the derived formulas.

Main Results:

• The proposed formulas provide accurate power estimations comparable to simulation results for various correlated hierarchical endpoints.
• The method allows for the evaluation of power based on the number, ordering, and types of endpoints.
• Formulas are effective across different types of hierarchical endpoints, with slight discrepancies in very high correlation scenarios.

Conclusions:

• The developed formulas offer a practical and robust alternative to simulation-based calculations for win measures in hierarchical endpoint analysis.
• These formulas facilitate meaningful and justifiable specification of win measures and tie probabilities.
• The approach enhances the efficiency and accuracy of sample size and power calculations in clinical trial design.