Optimal study designs for cluster randomised trials: An overview of methods and results

Area of Science:

• Biostatistics
• Clinical Trial Design

Background:

• Cluster randomized trials (CRTs) offer efficient designs for studies where randomization by individual is impractical.
• Optimizing CRT designs is challenging due to intra-cluster correlation and temporal dependencies in observations.

Approach:

• This article reviews statistical and computational methods for identifying optimal cluster randomized trial designs.
• Methods adapted from experimental design for correlated observations are applied to the CRT context.
• Three classes of methods are presented: exact formulae for variance, generalized weighting methods, and combinatorial optimization algorithms.

Key Points:

• Methods include deriving algorithms/weights from variance formulae, generalized experimental unit weighting, and combinatorial subset selection.
• Discussion covers rounding weights, extensions to non-Gaussian models, and robust optimality.
• Examples demonstrate optimal cluster allocation and observation number determination for Gaussian/non-Gaussian models with various covariance structures.

Conclusions:

• The presented methods provide a framework for optimizing cluster randomized trial designs.
• Efficient design selection can enhance statistical power and resource allocation in cluster randomized trials.