A simple sample size formula for analysis of covariance in cluster randomized trials

Area of Science:

• Biostatistics
• Clinical Trials Methodology

Background:

• Sample size calculations for cluster randomized trials often rely solely on follow-up outcome scores.
• Baseline outcome measurements are frequently available or obtainable, offering potential for more powerful analyses.

Purpose of the Study:

• To evaluate the efficiency of using analysis of covariance (ANCOVA) with baseline and follow-up scores compared to using follow-up scores alone.
• To determine how baseline measurements impact sample size requirements in cluster randomized trials.

Main Methods:

• Calculated the efficiency gain of ANCOVA by comparing it to an analysis using only follow-up scores.
• Expressed the reduction in sample size as a factor of r², where r is the correlation between cluster means at baseline and follow-up.
• Related the cluster correlation (r) to subject and cluster autocorrelation parameters.

Main Results:

• ANCOVA using baseline and follow-up scores is more powerful than using follow-up scores alone for continuous outcomes.
• The required sample size for ANCOVA is reduced by a factor of r², where r is the correlation of cluster means between baseline and follow-up.
• Subject and cluster autocorrelation influence the efficiency gain, potentially leading to substantial reductions in the number of clusters needed.

Conclusions:

• Incorporating baseline measurements via ANCOVA offers a statistically more efficient approach for cluster randomized trials.
• Subject matter expertise can inform plausible values for autocorrelation parameters, aiding sample size calculations when prior estimates are unavailable.
• Analysis of covariance can significantly decrease the number of clusters required, optimizing trial resource allocation.