The Use of a Correlated Binary Data Model for Sample Size Calculation and Analysis of Accessions Data

• Activity Number: 138
• Type: Topic Contributed
• Date/Time: Monday, August 7, 2006 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): John C. Evans*+ and Corina M. Sirbu and Kay M. Larholt and Brian Johnson
• Companies: Boston Scientific Corporation and Boston Scientific Corporation and Boston Scientific Corporation and Boston Scientific Corporation
• Keywords: implantable port venous accessions ; complication rate ; temporally correlated binary data ; sample size ; interim analysis

Abstract

Sample size calculations and analyses in studies of port comparisons have assumed independence of accessions across time. However, accessions usually are correlated, thus requiring more study patients. If this correlation is ignored when counting complications for the individual accessions, and the proportions of complications are compared by methods assuming independence of accessions, then incorrect conclusions may be drawn. We will present results for a new study, comparing sample size calculations and corresponding interim analyses under the usual assumption of independence of accessions with those obtained when a positive correlation is assumed among accessions. In the latter case, we use a correlated binary data model to estimate the number of patients required with an average of 13 accessions over a six months and do a corresponding interim analysis.