Article type
Abstract
"Background: Response adaptive randomization (RAR) adjusts the randomization probability based on the treatment outcomes during the trial process so that more subjects are treated with better therapies and can benefit from the clinical trial, which conforms to ethical requirements.
Objectives: This study aimed to review research progress of RAR methods, and outline the details of a RAR platform developed for randomized controlled trials (RCTs).
Methods: We retrieved articles in PubMed, Embase, Cochrane library and Web of Science databases from their inception to January 2023 to summarize the origin, philosophy, classification, development, algorithm and application of the RAR methods. Doubly adaptive biased coin design (DBCD) is one of the robust algorithms. We developed a RAR platform for clinical research based on the DBCD method.
Results: Based on the comprehensive and systematic literature retrieval, the research progress on RAR methods was summarized. The RAR methods are mainly divided into parameter and non-parameter methods. Parameter methods usually target optimal allocation, including DBCD and sequential maximum likelihood procedure. During the first stage, a number of subjects are randomly enrolled with equal probability, and the unknown parameters are estimated according to the accrual of outcomes. The next subject will be assigned with the updated allocation proportion determined by the allocation function. Non-parameter methods contain the generalized Friedman’s urn model, play-the-winner rule, drop-the-loser rule etc.. Non-parameter method could adjust the allocation probability among various treatments according to the predefined rules, but fail to target optimal allocation. The RAR platform has been developed, which integrates data management, the DBCD algorithm and drug logistics.
Conclusions: This study reviewed the latest research progress on RAR methods, introduced characteristics and rules of parameter and non-parameter methods, particularly DBCD, and summarized their clinical application. The RAR platform has been developed, which can provide technical and methodological supports for RCTs, by reducing sample size, enabling more subjects to receive superior interventions, and satisfying ethical requirements. A demo and trial version of this platform were also provided. "
Objectives: This study aimed to review research progress of RAR methods, and outline the details of a RAR platform developed for randomized controlled trials (RCTs).
Methods: We retrieved articles in PubMed, Embase, Cochrane library and Web of Science databases from their inception to January 2023 to summarize the origin, philosophy, classification, development, algorithm and application of the RAR methods. Doubly adaptive biased coin design (DBCD) is one of the robust algorithms. We developed a RAR platform for clinical research based on the DBCD method.
Results: Based on the comprehensive and systematic literature retrieval, the research progress on RAR methods was summarized. The RAR methods are mainly divided into parameter and non-parameter methods. Parameter methods usually target optimal allocation, including DBCD and sequential maximum likelihood procedure. During the first stage, a number of subjects are randomly enrolled with equal probability, and the unknown parameters are estimated according to the accrual of outcomes. The next subject will be assigned with the updated allocation proportion determined by the allocation function. Non-parameter methods contain the generalized Friedman’s urn model, play-the-winner rule, drop-the-loser rule etc.. Non-parameter method could adjust the allocation probability among various treatments according to the predefined rules, but fail to target optimal allocation. The RAR platform has been developed, which integrates data management, the DBCD algorithm and drug logistics.
Conclusions: This study reviewed the latest research progress on RAR methods, introduced characteristics and rules of parameter and non-parameter methods, particularly DBCD, and summarized their clinical application. The RAR platform has been developed, which can provide technical and methodological supports for RCTs, by reducing sample size, enabling more subjects to receive superior interventions, and satisfying ethical requirements. A demo and trial version of this platform were also provided. "