Statistical inference for extended or shortened phase II studies based on Simon's two-stage designs

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Oncology Research

Background:

• Simon's two-stage designs are widely used in Phase II oncology trials to minimize patient exposure to ineffective treatments.
• Existing inference procedures often overlook the actual sampling plan, potentially leading to inaccurate results.
• Koyama and Chen (2008) proposed an inference method addressing variations in second-stage sample sizes.

Purpose of the Study:

• To develop and evaluate an alternative inference method for Simon's two-stage designs.
• To address limitations of existing methods, particularly when actual sample sizes deviate from planned ones.
• To provide more accurate statistical inference for adaptive clinical trial designs.

Main Methods:

• An alternative inference method based on the likelihood ratio is proposed.
• Permissible sample paths under Simon's designs are ordered by their conditional likelihood.
• P-values are calculated using the standard definition: the probability of observing a test statistic as extreme or more extreme than the observed one under the null hypothesis.

Main Results:

• The proposed method offers inference in scenarios where Koyama and Chen's method is challenging to apply.
• Numerical simulations indicate the new method yields smaller biases and narrower confidence intervals with similar coverage rates.
• The method's performance was demonstrated using a real-world data example, comparing it with existing approaches.

Conclusions:

• Inference procedures for Simon's designs frequently disregard the actual sampling plan, leading to unadjusted P-values, point estimates, and confidence intervals.
• The adaptiveness of the design is often not accounted for in standard statistical reporting.
• The study emphasizes the necessity of employing proper statistical inference procedures for accurate trial interpretation.