Two Questions About the Design of Cluster Randomized Trials: A Tutorial

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Public Health Research

Background:

• Cluster randomized trials (CRTs) are employed when interventions must be applied at a group level or due to logistical considerations like feasibility and contamination.
• CRTs present statistical and logistical challenges compared to individually randomized trials, necessitating careful consideration of their rationale.

Purpose of the Study:

• To provide guidance on determining the appropriateness of using cluster randomized trials (CRTs).
• To outline methods for calculating sample sizes for CRTs.
• To illustrate these concepts using an ongoing CRT in intensive care units.

Main Methods:

• Discusses the criteria for selecting CRTs over individually randomized trials.
• Explains sample size calculation using the "design effect" to quantify within-cluster correlation.
• Highlights the trade-offs between the number and size of clusters regarding statistical efficiency and feasibility.

Main Results:

• CRTs are indicated when group-level interventions are essential or practical.
• Sample size calculation requires inflating individually randomized trial calculations by the "design effect".
• Balancing cluster number and size is crucial for statistical power and feasibility, with sensitivity analyses recommended.

Conclusions:

• The decision to conduct a CRT requires careful evaluation of necessity and feasibility.
• Accurate sample size determination in CRTs is critical and involves accounting for the "design effect".
• Collaboration with statisticians and conducting sensitivity analyses are essential for robust CRT design and sample size calculation.