Planning a cluster randomized trial with unequal cluster sizes
Author Information
Author(s): Guittet Lydia, Ravaud Philippe, Giraudeau Bruno
Primary Institution: Département d'Epidémiologie, Biostatistique et Recherche Clinique, Groupe Hospitalier Bichat-Claude Bernard (AP-HP), Université Xavier Bichat, Paris, France
Hypothesis
How does an imbalance in cluster size affect the power of cluster randomized trials?
Conclusion
Imbalance in cluster sizes can significantly influence the power of trials, especially when the number of clusters is low and/or the intraclass correlation coefficient is high.
Supporting Evidence
- Imbalance in cluster sizes can lead to underpowered trials.
- Minimum variance weights correction is recommended for sample size adjustments.
- Severe imbalances, like the Pareto imbalance, should be accounted for in trial design.
Takeaway
When planning a study where groups are made up of clusters, it's important to consider that some clusters might be much larger than others, which can affect the study's results.
Methodology
Simulations were performed to study the impact of cluster size imbalance on power and to evaluate methods for adjusting sample size calculations.
Potential Biases
Potential bias due to unequal recruitment across clusters was not fully addressed.
Limitations
The study focused on normally distributed continuous outcomes and did not account for differential recruitment between arms.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website