Efficient group sequential designs when there are several effect sizes under consideration

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methods

Background:

• Group sequential designs are crucial for efficient clinical trials, balancing statistical rigor with resource optimization.
• Standard and adaptive designs aim to maintain Type I error and power while controlling sample size.
• Optimizing expected sample size across a range of effect sizes is a key challenge in trial design.

Purpose of the Study:

• To construct efficient group sequential designs with low expected sample size at the null, alternative, and optimistic effect sizes.
• To compare the performance of four nested classes of group sequential and adaptive designs.
• To identify simple yet efficient designs suitable for practical application in clinical research.

Main Methods:

• Investigated four nested classes of designs: rho-family group sequential (A), unequal initial group size (B), fixed arbitrary boundaries (C), and adaptive designs (D).
• Evaluated optimal procedures within each class based on expected sample size, Type I error rate, and power.
• Utilized tables and figures to illustrate design performance and define optimal procedures.

Main Results:

• Class B designs, featuring unequal initial group sizes, demonstrated high efficiency.
• The performance of Class B designs closely approached that of more complex Class C and D designs.
• Optimal procedures within Class A and B were identified and illustrated.

Conclusions:

• Class B group sequential designs offer a compelling combination of simplicity and efficiency.
• These designs provide a practical alternative to more complex adaptive procedures without significant loss in efficiency.
• The findings support the use of Class B designs for constructing efficient clinical trials across various anticipated effect sizes.