Using Power Analysis to Choose the Unit of Randomization, Outcome, and Approach for Subgroup Analysis for a...

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Cardiovascular Disease Prevention

Background:

• Randomized controlled trials (RCTs) are essential for evaluating interventions but require significant resources.
• Optimizing RCT design can enhance statistical power, potentially reducing sample size and associated costs.
• Multilevel study designs introduce complexities in power calculations and analysis.

Purpose of the Study:

• To demonstrate design features in randomized controlled trials that increase statistical power and decrease sample size and costs.
• To illustrate these features using a proposed multilevel cardiovascular prevention research study.
• To provide example code for power analyses using validated software.

Main Methods:

• Examined three design features: level of randomization in multilevel trials, use of multivariate vs. composite outcomes, and pooled vs. stratified analysis.
• Calculated power using published, exact analytic methods for continuous outcomes.
• Applied methods to a proposed RCT randomizing adults to telehealth or in-person treatment for cardiovascular risk reduction, measuring Essential Eight scores.

Main Results:

• Power varies significantly with the chosen level of randomization in multilevel trials.
• Testing multivariate outcomes improves power and interpretability compared to unweighted composite outcomes.
• A pooled analytic approach offers greater power for intervention effect testing than stratified analysis.

Conclusions:

• Strategic design choices in RCTs, including randomization level, outcome variable selection, and analysis strategy, can significantly improve statistical power.
• These optimizations are crucial for efficient and cost-effective clinical research, particularly in complex multilevel studies.
• The findings support the proposed cardiovascular prevention trial's design and offer a framework for future research.