A mixed model-based variance estimator for marginal model analyses of cluster randomized trials

Area of Science:

• Biostatistics
• Clinical Trials Methodology

Background:

• Generalized estimating equations (GEE) are widely used for cluster randomized trials (CRTs) due to their population-averaged interpretation and software availability.
• However, GEE often underestimates standard errors in CRTs, potentially leading to inaccurate conclusions.
• Penalized quasi-likelihood (PQL) offers better standard error estimation but provides cluster-specific estimates and lacks widespread software implementation.

Purpose of the Study:

• To develop an improved variance estimator for GEE in CRT analysis.
• To enhance the utility of GEE for CRTs by addressing standard error underestimation.
• To propose a variance estimator that can be integrated into existing GEE software.

Main Methods:

• A novel sandwich-type variance estimator was derived by adapting the PQL variance estimator.
• The proposed estimator was evaluated using numerical examples and real data from a CRT.
• Performance was compared against existing variance estimators in the literature.

Main Results:

• The proposed variance estimator demonstrated comparable or superior performance to existing methods.
• The new estimator effectively addresses the standard error underestimation issue in GEE for CRTs.
• The approach facilitates the incorporation of improved variance estimation into standard GEE packages.

Conclusions:

• The proposed sandwich-type variance estimator offers a practical solution for improving GEE analysis in CRTs.
• This method enhances the accuracy of intervention effect estimates in CRTs.
• The findings support the broader adoption of GEE with this improved variance estimation technique for CRTs.