Blinded sample size re-estimation in three-arm trials with 'gold standard' design

Area of Science:

• Biostatistics
• Clinical Trial Design

Background:

• The 'gold standard' design uses three arms: experimental treatment, active control, and placebo.
• Assessing non-inferiority and assay sensitivity in three-arm trials involves pairwise comparisons.
• Blinded sample size re-estimation is crucial for maintaining trial power.

Purpose of the Study:

• To evaluate blinded sample size re-estimation procedures in the 'gold standard' design.
• To identify optimal methods for normally distributed outcomes using an absolute margin approach.
• To propose a novel approach to overcome underpowered trials resulting from certain re-estimation methods.

Main Methods:

• A simulation study was conducted to assess operating characteristics (power, type I error).
• Comparison of sample size re-estimation procedures using common and unbiased variance estimators.
• Development and testing of an inflation factor for the Xing-Ganju variance estimator.

Main Results:

• Re-estimation using the one-sample variance estimator leads to overpowered trials.
• Re-estimation with unbiased estimators like Xing-Ganju results in underpowered trials.
• The proposed inflation factor with the Xing-Ganju estimator yields adequately powered trials.

Conclusions:

• The proposed inflation factor for the Xing-Ganju variance estimator provides a robust solution for sample size re-estimation.
• This method ensures adequate power without introducing bias into effect estimates.
• The inflation factor can be pre-calculated, simplifying trial conduct.