Sequential Multiple Assignment Randomized Trials Based on Restricted Mean Survival Time

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Survival Analysis

Background:

• Sequential Multiple Assignment Randomized Trials (SMART) are complex for adaptive treatment strategies (ATSs), especially with survival data.
• Existing methods face analytical challenges and resource intensity due to long follow-ups and proportional hazards assumption violations.

Purpose of the Study:

• To propose a novel statistical inference framework for SMART designs with survival outcomes.
• To address the analytical complexity and resource demands of adaptive clinical trials.

Main Methods:

• Utilized Restricted Mean Survival Time (RMST) as a robust summary measure, independent of the proportional hazards assumption.
• Developed fixed- and dynamic-weight RMST estimators, including variance-covariance structures, confidence intervals, and hypothesis tests.
• Integrated interim analyses into SMART designs with RMST and implemented a type I error control method.

Main Results:

• The proposed RMST framework demonstrated good estimator performance in simulations.
• Interim analyses integrated with RMST can effectively reduce sample size or trial duration.
• The framework provides efficient and practical tools for comparing adaptive treatment strategies.

Conclusions:

• The RMST-based framework offers an efficient and practical solution for SMART trials involving survival outcomes.
• This approach simplifies the analysis of adaptive treatment strategies, enhancing personalized medicine research.
• The study provides robust statistical methods for adaptive clinical trial designs.