The effect of cluster size variability on statistical power in cluster-randomized trials

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Epidemiology

Background:

• Cluster-randomized trials (CRTs) are essential for interventions not feasible at the individual level.
• CRT design and analysis present complexities compared to individually randomized trials.
• Accurate sample size calculation, including cluster number, is crucial for CRT statistical power.

Purpose of the Study:

• To investigate the impact of variable cluster sizes on the statistical power of CRTs.
• To evaluate how deviations from a fixed mean cluster size assumption affect trial power.

Main Methods:

• A simulation study was conducted to assess statistical power under varying cluster sizes.
• The simulation examined the relationship between cluster size variability and power.

Main Results:

• Increased variability in cluster size generally leads to a reduction in statistical power for CRTs.
• This effect highlights a potential pitfall in sample size calculations that assume fixed cluster sizes.

Conclusions:

• Researchers must account for potential cluster size variation when designing CRTs.
• Failure to consider cluster size variability may lead to underpowered studies and inaccurate conclusions.