Sequential urn designs with elimination for comparing K > or =3 treatments

Area of Science:

• Biostatistics
• Clinical Trial Design
• Medical Research Methodology

Background:

• Comparing multiple treatments (K >= 3) with binary outcomes requires efficient adaptive designs.
• Sequential procedures can optimize patient allocation and treatment selection during trials.
• Existing methods may not fully leverage adaptive randomization for early elimination of ineffective treatments.

Purpose of the Study:

• To propose and evaluate a fully sequential procedure for comparing K >= 3 treatments with binary responses.
• To incorporate adaptive urn designs and stopping rules for efficient treatment elimination.
• To assess the performance of the proposed procedure, including treatment failures and allocation, and its application in dose-finding studies.

Main Methods:

• A fully sequential procedure employing adaptive urn designs for patient randomization.
• Incorporation of stopping rules to eliminate less promising treatments during the trial.
• Simulation studies to evaluate the performance of various adaptive urn designs.
• Application and comparison in a lung cancer study and dose-finding scenarios.

Main Results:

• The 'drop-the-loser' rule demonstrated superior effectiveness compared to equal allocation and other designs.
• The sequential elimination procedure efficiently allocates patients and reduces treatment failures.
• The method is practical for real-world applications, as shown in a lung cancer study.
• Performance comparison with a recent method in dose-finding studies indicates its viability.

Conclusions:

• The proposed sequential procedure with adaptive urn designs and the 'drop-the-loser' rule is effective for comparing multiple treatments.
• This approach enhances efficiency in clinical trials by enabling early elimination of inferior treatments.
• The methodology offers practical benefits and potential for extensions in dose-finding and other adaptive trial designs.