Covariate-adjusted response-adaptive designs for semiparametric survival models.

Area of Science:

• Clinical Trials
• Biostatistics
• Survival Analysis

Background:

• Covariate-adjusted response adaptive (CARA) designs aim to maximize patient benefit in clinical trials.
• Existing CARA designs often rely on parametric assumptions, limiting their real-world applicability.
• Accelerated failure time (AFT) models offer valid inference for some CARA designs but are rarely used in primary analyses.

Purpose of the Study:

• To develop novel CARA designs for survival trials that do not require distributional assumptions.
• To ensure valid statistical inference under the proportional hazards assumption.
• To implement an optimal allocation approach for multiple experimental objectives.

Main Methods:

• Proposed CARA designs obviate distributional assumptions, relying on the proportional hazards assumption.
• Utilized covariate-adjusted doubly adaptive biased coin and covariate-adjusted efficient-randomized adaptive designs for patient randomization.
• Employed sequential estimation of Cox regression coefficients to achieve optimal allocation targets.

Main Results:

• Extensive simulation studies demonstrated favorable operating characteristics of the proposed designs.
• The designs were successfully implemented to re-design a real-life confirmatory clinical trial.
• The new CARA designs provide valid statistical inference without parametric assumptions.

Conclusions:

• The developed CARA designs offer a flexible and robust approach for adaptive clinical trials involving survival outcomes.
• These methods enhance the efficiency of clinical trials by optimizing treatment allocation.
• The findings support the broader adoption of assumption-free adaptive designs in survival trial settings.