Sample size in cluster-randomized trials with time to event as the primary endpoint

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Epidemiology

Background:

• Cluster-randomized trials (CRTs) involve randomizing groups to interventions.
• Time-to-event data analysis is crucial for many CRTs.
• Correlations within clusters impact statistical power and require specific sample size considerations.

Purpose of the Study:

• To derive a practical sample size formula for clustered time-to-event data.
• To extend existing sample size calculations to account for clustered designs.
• To provide a tool for planning CRTs with correlated event times.

Main Methods:

• Development of a sample size formula based on shared frailty models for clustered time-to-event data.
• Assumed constant marginal baseline hazards and intra-cluster correlation.
• Validation through simulations and illustration with a real-world trial.

Main Results:

• A novel, easy-to-apply sample size formula for CRTs with time-to-event outcomes was derived.
• The formula extends Schoenfeld's formula by incorporating cluster correlations.
• Simulations confirmed the formula's accuracy, even with non-constant baseline hazards.

Conclusions:

• The derived formula is a valuable tool for sample size determination in CRTs with time-to-event data.
• Accurate sample size planning is essential for detecting intervention effects in clustered studies.
• The methodology is applicable to various health research settings, including quality improvement initiatives.