An evaluation of increasing sample size based on conditional power.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methods

Background:

• Sample size re-estimation (SSR) is crucial for adaptive clinical trial designs.
• Evaluating SSR designs, like the promising zone design, is essential for optimizing trial efficiency.
• Measures of SSR performance include probability of sample size increase, conditional power, and expected sample size/power gains.

Purpose of the Study:

• To evaluate sample size re-estimation (SSR) designs, comparing late-stage versus mid-study re-estimations.
• To assess the impact of true effect sizes on SSR performance.
• To determine the optimal timing for SSR to maximize statistical power and resource efficiency.

Main Methods:

• Simulation studies evaluating SSR designs under varying true effect sizes (0.4 to 1.1 of protocol-specified effect size).
• Analysis of six key performance measures: probability of sample size increase, mean proportional increase, conditional power (with/without increase), and expected increase in sample size and power.
• Comparison of late-stage versus mid-study SSR timing.

Main Results:

• Late-stage SSR demonstrates clear superiority over mid-study SSR.
• Mid-study SSR can be inefficient, especially when the true effect size is near the protocol-specified size or of limited clinical importance.
• Late-stage SSR results in smaller, more targeted studies with a greater increase in overall power.

Conclusions:

• Conducting SSR late in a study is the optimal strategy for enhancing statistical power and trial efficiency.
• Mid-study SSR is generally inefficient and may lead to unnecessary sample size adjustments.
• Timely SSR, particularly when performed late, improves the precision and impact of clinical trial outcomes.