Adaptive adjustment of the randomization ratio using historical control data.

Area of Science:

• Clinical Trials
• Biostatistics
• Oncology

Background:

• Prospective trials often use historical control data retrospectively.
• This data is typically used for population-averaged effect estimation in meta-analyses.

Purpose of the Study:

• Investigate an adaptive trial design for a colorectal cancer trial.
• Utilize historical control data to balance information between concurrent and historical arms.

Main Methods:

• Implement adaptive randomization to balance total information (concurrent and historical).
• Use effective historical sample size (EHSS) to guide allocation probabilities.
• Employ commensurate priors for interim heterogeneity assessment and final analysis borrowing from historical data.

Main Results:

• Adaptive randomization assigns more patients to novel therapy when historical controls are unbiased.
• Bias in historical controls leads to more equal patient allocation.
• The commensurate prior model yields admissible estimators with good bias-variance trade-offs.

Conclusions:

• Adaptive randomization is sensitive to population drift but suitable for gradual enrollment.
• Balancing information in time-to-event analyses is challenging with informative right-censoring.
• The proposed design is valuable for trials following prior control therapy evaluations, especially when historical data reliance is necessary.