Incorporating Contact Network Structure in Cluster Randomized Trials

Area of Science:

• Epidemiology
• Biostatistics
• Clinical Trial Design

Background:

• Randomized controlled trials (RCTs) are standard for evaluating treatment efficacy.
• Cluster randomized trials (CRTs) are often used for infectious diseases, assigning groups (clusters) to interventions.
• Trial power, the ability to detect a true effect, can be compromised by cluster structure and interactions.

Purpose of the Study:

• To investigate the impact of within- and between-cluster interactions on statistical power in infectious disease CRTs.
• To compare simulation-based power assessments with traditional formula-based methods.
• To highlight the importance of accounting for network structure and cross-contamination in trial design.

Main Methods:

• Simulated an infectious process across a network of clusters.
• Varied parameters including within-cluster network structure, between-cluster mixing (cross-contamination), and infectivity.
• Compared power calculations derived from simulations versus standard formulas.

Main Results:

• Formula-based power calculations can be overly conservative with low between-cluster mixing.
• Failure to account for moderate to high between-cluster mixing significantly reduces study power.
• Within-cluster network structure influences power, particularly for infections spreading through highly connected individuals.

Conclusions:

• Current formula-based power calculations may underestimate power in some scenarios and overestimate it in others if mixing is not considered.
• Simulation-based approaches incorporating network information are crucial for accurate power assessment in CRTs for infectious diseases.
• Empirical data on mixing patterns can improve the design and power estimations of future trials.