Design and analysis considerations for cohort stepped wedge cluster randomized trials with a decay correlation...

Area of Science:

• Biostatistics
• Clinical Trial Design
• Longitudinal Data Analysis

Background:

• Stepped wedge cluster randomized trials (SW-CRTs) are longitudinal designs where clusters sequentially adopt an intervention.
• Traditional SW-CRT analyses often simplify correlation structures, potentially leading to inaccurate sample size and power calculations.
• Existing methods rarely account for the decay of correlation between measurements over time within clusters.

Purpose of the Study:

• To propose a statistical framework for SW-CRTs that incorporates a proportional decay correlation structure.
• To develop accurate methods for estimating correlation parameters and the marginal intervention effect in cohort SW-CRTs.
• To create and validate sample size and power calculation procedures that account for correlation decay.

Main Methods:

• Developed a matrix-adjusted quasi-least squares (MAQLS) approach for parameter estimation.
• Incorporated a proportional decay correlation structure to model within-cluster correlations over time.
• Investigated the accuracy of the proposed power procedure using a simulation study with continuous outcomes.

Main Results:

• The MAQLS approach accurately estimates correlation parameters and intervention effects, even with a small number of clusters (as few as nine).
• The developed power procedure shows good agreement between empirical power and predicted power.
• Bias-corrected sandwich variance is crucial for reliable analysis when using MAQLS.

Conclusions:

• The proposed methods provide a more accurate way to analyze cohort SW-CRTs with decaying correlations.
• The new sample size and power procedures enhance the planning and efficiency of SW-CRT studies.
• This framework offers a robust statistical foundation for complex longitudinal cluster randomized trial designs.