On efficient two-stage adaptive designs for clinical trials with sample size adjustment.

Area of Science:

• Clinical Trial Design
• Biostatistics
• Statistical Inference

Background:

• Group sequential designs are underutilized in clinical trials with significant overruns.
• Fixed sample size designs are traditionally used, but adaptive designs offer sample size adjustments.
• Existing two-stage adaptive designs can suffer from serious inefficiencies.

Purpose of the Study:

• To propose a novel likelihood-based two-stage adaptive design.
• To improve the efficiency of adaptive clinical trial designs.
• To provide methods for sequential statistical inference and defend existing adaptive design principles.

Main Methods:

• Developing a likelihood-based two-stage adaptive design using cumulative conditional power.
• Comparing the proposed design with group sequential designs and existing two-stage adaptive designs.
• Deriving methods for sequential p-values, confidence intervals, and unbiased estimates.

Main Results:

• The proposed design demonstrates superior or equal performance compared to group sequential designs.
• The new approach offers uniform improvement over existing two-stage adaptive designs.
• The inefficiency claims against adaptive designs by Tsiatis and Mehta (2003) are shown to be logically flawed.

Conclusions:

• The proposed likelihood-based two-stage adaptive design is an efficient alternative for clinical trials.
• This work provides a strong defense for the principles of adaptive clinical trial designs.
• The findings offer improved statistical inference methods for adaptive trial settings.