Sample Size for Pre-Post Comparison with Intervention Delivered to Cluster

• Activity Number: 657
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Zhiying You*+ and Ming Ji and O. Dale Williams
• Companies: The University of Alabama at Birmingham and San Diego State University and Florida International University
• Keywords: sample size ; power ; cluster ; relative efficiency ; single arm study

Abstract

Single arm studies delivering intervention to clusters of subjects have been used to assess strategies for disease prevention and control of chronic disease, but sample size approach is not available. We propose sample size methods for pre-post comparison with intervention delivered to clusters. Assume the pre-post difference is normally distributed and the paired t-test is used to test the intervention effect, then the t statistic has a central and non-central t distribution under the null and alternative hypotheses, taking account clustering. For unequal cluster sizes, the relative efficiency (RE) of unequal versus equal cluster size is defined as the ratio of the noncentrality parameter of unequal versus equal cluster sizes. The RE is straight forward to calculate, easy to interpret and connects directly the required mean cluster size to the required sample size with equal sizes for assessing study power. We thus can calculate the required mean cluster size to achieve the same power as an equal cluster sizes trial. There exists a minimum required number of clusters for a pre-specified power, which provides a quick evaluation of feasibility as to achieving a statistical power.