Randomization-based analysis of covariance for inference in the sequential parallel comparison design.

Area of Science:

• Clinical Trial Design
• Biostatistics
• Psychiatric Research

Background:

• Sequential parallel comparison design (SPCD) is useful for studies with high placebo response rates, such as psychiatric clinical trials.
• Excluding placebo responders from the first period in the second period is a key feature of SPCD.
• Existing methods may not fully leverage all available data in such designs.

Purpose of the Study:

• To present a methodology for analyzing treatment differences in SPCD, considering both overall and specific subgroups.
• To introduce hypothesis testing and analysis of covariance methods based on randomization distributions.
• To provide a statistical framework for studies with significant placebo effects.

Main Methods:

• Developed a hypothesis testing method based on randomization distribution to control type I error without assumptions.
• Introduced randomization-based analysis of covariance (ANCOVA) to adjust for baseline values.
• Proposed methods applicable to both the randomized population and a simple random sample from a larger population.

Main Results:

• The proposed methods control type I error effectively.
• Simulation studies demonstrate the statistical properties of the developed methods.
• The methodology was illustrated using data from the ADAPT-A clinical trial.

Conclusions:

• SPCD offers a valuable approach for clinical trials with high placebo response rates.
• The presented statistical methodology provides a robust framework for analyzing such trials.
• The proposed randomization-based methods ensure valid statistical inference.