Sample size calculation for a stepped wedge trial.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Epidemiology

Background:

• Stepped wedge trials (SWTs) are a type of clustered randomized trial with unique statistical design and analysis complexities.
• Existing literature on sample size and power calculations for SWTs is less extensive compared to standard parallel or clustered randomized clinical trials (CRTs).
• Accurate sample size calculations are crucial for SWTs to ensure valid estimation of intervention effects.

Purpose of the Study:

• To critically review existing analytical methods for sample size and power calculations in SWTs.
• To highlight the assumptions, validity, and potential extensions of current methods.
• To propose and evaluate simulation-based methods as an alternative to overcome limitations of analytical formulae.

Main Methods:

• A comprehensive literature review of analytical methods for SWT sample size and power calculations.
• A simulation exercise comparing simulation-based computations with analytical methods.
• Assessment of the impact of varying parameters (intracluster correlation, outcome type, data design) on sample size and power.

Main Results:

• SWTs may require fewer clusters than CRTs when intracluster correlation is high (>0.1).
• Simulation-based methods validated well against analytical methods, producing similar sample size/power results.
• Failure to account for time effects can lead to a gross overestimation of study power in SWTs.

Conclusions:

• Simulation-based procedures provide a robust framework for SWT sample size and power calculations, particularly for complex study-specific features.
• SWTs can be more efficient than CRTs in certain scenarios, depending on intracluster correlation and cluster size.
• The choice of trial design involves a balance of statistical efficiency, cost, trial duration, and other practical considerations.