Sample Size Strategies for Cluster Randomized Trials with Multiple Groups

• Activity Number: 632
• Type: Contributed
• Date/Time: Thursday, August 4, 2011 : 8:30 AM to 10:20 AM
• Sponsor: International Chinese Statistical Association
• Author(s): Zhiying You*+ and O Dale Williams and David Redden and Immaculada Aban and Edmond Kabagambe and Hemant Tiwari and Gary Cutter
• Companies: University of Alabama at Birmingham and University of Alabama at Birmingham and University of Alabama at Birmingham and University of Alabama at Birmingham and University of Alabama at Birmingham and University of Alabama at Birmingham and University of Alabama at Birmingham
• Keywords: cluster randomized trials ; sample size ; power ; multiple groups ; analysis of variance

Abstract

Sample size estimation for cluster randomized trials (CRTs) has focused mainly on comparing two intervention groups. Further, variable cluster sizes are common and this variation should be taken into account in sample size estimation. In this paper, we propose two sample size strategies for CRTs with comparison of multiple groups. The first strategy is based on pair-comparison of two groups with Bonferroni correction of Type I error rate and the second on analysis of variance (ANOVA). For the second strategy, we focus on trials with equal cluster sizes and then extend to variable cluster sizes by using an assessment of relative efficiency (RE) of unequal versus equal cluster sizes, which is defined based on non-centrality parameters in the setting of ANOVA. With variable cluster sizes, we construct sample size configure of an unequal cluster size trial that provide the same statistical power as an equal cluster size trial and accordingly our sample size approach may adjust the mean cluster size alone or both the number of clusters and the means cluster size simultaneously. The two strategies together provide sample size tool options for design with multiple intervention groups.