Statistical inference following sample size adjustment based on the 50%-conditional-power principle.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmaceutical Research

Background:

• Sample size adjustments in interim analyses can prevent study failure from low treatment effects.
• Conventional methods inflate Type I error rates when increasing sample size with null interim results.
• Resource efficiency is crucial, avoiding investment in ineffective drug candidates.

Purpose of the Study:

• To assess statistical inference following sample size adjustment based on promising interim results.
• To evaluate the 50% conditional power principle for adaptive clinical trial designs.
• To determine the validity of conventional statistical methods under this adjustment strategy.

Main Methods:

• Numerical studies were conducted to evaluate bias and coverage error.
• Analysis focused on maximum likelihood estimates (MLE) and confidence intervals.
• The 50% conditional power principle was applied for sample size increases.

Main Results:

• Bias in conventional maximum likelihood estimates (MLE) was generally small.
• Coverage error of conventional confidence intervals was also minimal.
• The 50% conditional power principle allows sample size increases without inflating Type I error rates.

Conclusions:

• Conventional, MLE-based statistical inference is recommended with the 50% conditional power principle.
• This approach ensures consistent statistics for hypothesis testing and inference.
• Adaptive trial designs can be effectively implemented using this sample size adjustment strategy.