Optimal multistage designs for randomised clinical trials with continuous outcomes

Area of Science:

• Clinical Trials
• Biostatistics
• Statistical Methodology

Background:

• Multistage clinical trial designs offer significant reductions in expected sample size.
• Stopping for futility or efficacy at interim stages is a key feature.
• The δ-minimax design aims to minimize the maximum expected sample size while adhering to error constraints.

Purpose of the Study:

• To identify δ-minimax designs for more than two stages, overcoming previous computational limitations.
• To compare these novel multistage designs with existing optimal designs and the triangular design.
• To explore admissible designs that balance maximum expected and maximum sample sizes.

Main Methods:

• Application of simulated annealing to identify δ-minimax designs for multistage trials.
• Comparison of δ-minimax designs with other optimal multistage and triangular designs.
• Utilizing the concept of admissible designs to balance sample size metrics.

Main Results:

• The δ-minimax design demonstrates good expected sample size properties across various treatment effects.
• However, it generally results in a higher maximum sample size compared to other designs.
• Admissible designs effectively balance maximum expected sample size with a reasonable maximum sample size.

Conclusions:

• The δ-minimax design is a valuable tool for optimizing clinical trial sample sizes.
• Admissible designs offer a practical approach for clinical trials by balancing efficiency and maximum sample size.
• These optimized multistage designs are highly appealing for future clinical trial applications.