Comparing different ways of calculating sample size for two independent means: A worked example

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methods

Background:

• Accurate sample size calculation is critical for the design of randomized controlled trials (RCTs).
• Different methods exist for sample size determination, influencing study power and resource allocation.
• Understanding these methods is essential for researchers planning clinical trials.

Purpose of the Study:

• To compare methods for sample size calculation in parallel two-arm RCTs.
• To provide insights into sample size determination at the design stage.
• To evaluate the impact of outcome measure choice and statistical methods on sample size requirements.

Main Methods:

• Comparison of sample size calculations using published RCT data.
• Evaluation of outcome measures: post-intervention score vs. change from baseline.
• Assessment of statistical methods: t-test vs. analysis of covariance (ANCOVA).
• Simulation studies to analyze the influence of correlation strength on sample size.

Main Results:

• Required sample size varies based on the outcome measure (post-intervention vs. change from baseline) and covariate inclusion.
• Simplified sample size equations have underlying assumptions that impact planning.
• Reporting mean change (SE) facilitates variance calculation, pooling, and correlation estimation for improved sample size determination.

Conclusions:

• The choice of primary outcome and statistical approach significantly affects sample size calculations in RCTs.
• Researchers must be aware of assumptions in simplified sample size formulas.
• Publishing mean change (SE) is strongly recommended to enable robust sample size calculations for future studies.