Sequential covariate-adjusted randomization via hierarchically minimizing Mahalanobis distance and marginal imbalance

Area of Science:

• Biostatistics
• Clinical Trial Design
• Comparative Studies

Background:

• Adaptive randomization methods are crucial for covariate balance in comparative studies.
• Existing methods often require paired or group allocations, posing practical challenges.
• Individual sequential allocation is needed to meet clinical demands without delay.

Purpose of the Study:

• To propose an individual and sequential patient randomization method.
• To address limitations of existing group/pair allocation schemes in clinical trials.
• To enhance control over randomization by balancing covariates and group sizes.

Main Methods:

• Developed a novel sequential randomization method for individual patient allocation.
• Introduced a modified Mahalanobis distance to measure covariate imbalance.
• Conceptually separated and prioritized minimization of covariate and marginal imbalance.

Main Results:

• The proposed method achieves optimal covariate balance.
• It effectively maintains marginal balance (sample size difference) directly.
• Demonstrated superior performance via simulations and real data analysis.

Conclusions:

• The new method offers superior control over randomization compared to existing techniques.
• Provides theoretical guarantees for imbalance measure convergence and treatment effect estimation.
• Facilitates more efficient and balanced clinical trial designs.