Confidence intervals for odds ratio from multistage randomized phase II trials.

Area of Science:

• Clinical Trials
• Biostatistics
• Medical Research

Background:

• Multi-stage randomized trial designs offer enhanced efficiency through early termination based on efficacy signals.
• Standard statistical inference methods can be invalidated by the sequential nature of multi-stage trials.
• Accurate analysis is crucial for reliable conclusions in adaptive clinical trial designs.

Purpose of the Study:

• To develop valid statistical inference methods for multi-stage randomized phase II trials.
• To propose exact conditional confidence intervals for the odds ratio with dichotomous outcomes.
• To address the invalidity of traditional single-stage confidence intervals in adaptive trial settings.

Main Methods:

• Focus on multi-stage randomized phase II trials with dichotomous outcomes (e.g., treatment response).
• Propose exact conditional confidence intervals for the odds ratio.
• Utilize a linear ordering of outcomes conditioned on the total number of responders per stage and the exact conditional distribution function.

Main Results:

• Developed a method for valid statistical inference in multi-stage randomized trials.
• Proposed exact conditional confidence intervals that account for the trial's multi-stage structure.
• Demonstrated the invalidity of standard single-stage confidence intervals in this context.

Conclusions:

• The proposed method provides accurate confidence intervals for odds ratios in multi-stage phase II trials.
• Exact conditional inference is essential for maintaining statistical validity in adaptive trial designs.
• This approach enhances the reliability of results from efficient, early-terminating clinical trials.