On sample size and inference for two-stage adaptive designs

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• Adaptive designs are crucial for maintaining Type I error rates in clinical trials.
• Practical adaptive designs must balance statistical power with sample size limitations.
• Existing two-stage designs may not fully address power and sample size constraints simultaneously.

Purpose of the Study:

• To propose a novel two-stage adaptive design that achieves desired statistical power.
• To limit the maximum overall sample size in clinical trials.
• To develop robust statistical inference methods for two-stage adaptive designs.

Main Methods:

• The proposed design incorporates a main stage and an extension stage.
• The main stage is powered for anticipated effect sizes.
• The extension stage allows sample size augmentation if the true effect size is smaller than anticipated.
• Methods for calculating overall adjusted p-values, point estimates, and confidence intervals are developed.
• An exact two-stage test procedure is outlined for enhanced robustness.

Main Results:

• The proposed two-stage adaptive design maintains the Type I error rate.
• The design allows for sample size adjustments to achieve desired statistical power.
• The methods developed provide accurate statistical inference, including adjusted p-values and confidence intervals.
• The exact two-stage test procedure ensures robust inference.

Conclusions:

• The proposed two-stage adaptive design offers a practical solution for clinical trials.
• It effectively balances statistical power and sample size constraints.
• The developed inference methods support reliable decision-making in adaptive trial settings.