Boundary problem in Simon's two-stage clinical trial designs.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• Assessing new treatments in clinical trials with binary endpoints involves comparing observed response rates to target rates.
• Traditional statistical inference uses one-sided hypotheses, requiring Type I error rate computation across the null hypothesis parameter space.
• The monotonicity property is essential for ensuring the Type I error rate occurs at the boundary.

Purpose of the Study:

• To theoretically prove the monotonicity property for one-arm two-stage clinical trial designs.
• To establish conditions under which this critical statistical property holds.
• To contribute towards a complete future proof of the monotonicity property.

Main Methods:

• Theoretical mathematical proof for the monotonicity property under specific design conditions.
• Numerical simulations to verify the property across a range of sample sizes.
• Analysis of one-arm two-stage designs.

Main Results:

• The monotonicity property was theoretically proven for designs where the final threshold is less than the first-stage sample size, given a weak additional condition.
• Numerical evidence confirmed the monotonicity property for designs with first and second-stage sample sizes ranging from 10 to 100.
• The study provides a foundational step towards a complete proof of this property.

Conclusions:

• The monotonicity property is demonstrated to hold for a class of one-arm two-stage designs.
• This finding has implications for the reliable statistical assessment of treatment efficacy in clinical trials.
• Further research may build upon these methods to achieve a comprehensive proof.