Group Response-Adaptive Randomization With Delayed and Missing Responses

Area of Science:

• Clinical Trials
• Biostatistics
• Medical Research Methodology

Background:

• Traditional response-adaptive randomization (RAR) updates assignments after each patient response, posing practical challenges in many clinical settings.
• Existing RAR methods often struggle with delayed or missing patient response data, limiting their real-world applicability.

Purpose of the Study:

• To propose a novel family of group response-adaptive randomization (RAR) procedures for clinical trials.
• To address the impracticality of individual response-based RAR by introducing group-based updates.
• To evaluate the performance and theoretical properties of the proposed group RAR design, particularly in scenarios with delayed or missing data.

Main Methods:

• Development of a new group response-adaptive randomization procedure updating assignments based on aggregated patient responses.
• Analysis of the proposed design's theoretical properties, comparing them to existing doubly adaptive biased coin designs (DBCD).
• Numerical simulations to assess performance under various conditions, including delayed and randomly missing responses.

Main Results:

• The proposed group doubly adaptive biased coin design (DBCD) maintains essential theoretical properties similar to existing DBCDs.
• The group RAR design demonstrates robust performance in simulations involving delayed and randomly missing patient responses.
• Successful application of the new group RAR design in a real clinical trial, showcasing its practicality and advantages.

Conclusions:

• The proposed group response-adaptive randomization offers a practical and effective alternative to individual-response-based RAR in clinical trials.
• This group RAR design is suitable for handling delayed and missing response data, enhancing trial efficiency.
• The findings support the broader application of group response-adaptive designs in clinical research.