Maintaining the validity of inference in small-sample stepped wedge cluster randomized trials with binary outcomes...

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Epidemiology

Background:

• Stepped wedge cluster trials (SWTs) are increasingly used, often with few clusters and multiple time intervals.
• Common analysis using generalized linear mixed models may violate assumptions due to the intraclass correlation structure.
• Alternative methods are needed to ensure valid inference in small-sample SWTs.

Purpose of the Study:

• To evaluate the validity of generalized estimating equations (GEE) for analyzing small-sample stepped wedge cluster trials with binary outcomes.
• To identify optimal bias correction methods for standard errors and degrees of freedom in GEE for SWTs.

Main Methods:

• Extensive simulation study based on a motivating example and a general design.
• Comparison of generalized linear mixed models with generalized estimating equations (GEE).
• Assessment of various bias correction techniques for standard error estimates and degrees of freedom.

Main Results:

• Generalized estimating equations (GEE) maintain the validity of inference in small-sample SWTs with binary outcomes.
• Specific combinations of bias corrections for standard errors and degrees of freedom were identified as optimal.
• The performance was evaluated based on attaining nominal type I error rates.

Conclusions:

• Generalized estimating equations (GEE) provide a robust and valid analytical approach for small-sample stepped wedge cluster trials (SWTs).
• Careful selection of bias correction methods for standard errors and degrees of freedom is crucial for accurate inference in SWTs.
• This research offers practical guidance for analyzing complex clinical trial designs with limited clusters.