A Model for Multilevel Clustered Data with Informative Cluster Size Using Poisson Distribution

• Activity Number: 256
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Ana-Maria Iosif*+ and Allan Sampson
• Companies: University of California, Davis and University of Pittsburgh
• Keywords: clustered data ; informative cluster size ; Poisson distribution

Abstract

We consider that clustered data are collected repeatedly (over conditions or over time) and we look on a way of modeling and analyzing repeated measures clustered data when the cluster size is informative. In a clinical trial of a migraine drug, the number of headaches and the degree of severity associated to each headache are collected repeatedly, over follow-up periods. Both the number of migraines and their severity levels are informative about the drug performance. A change in disease status is reflected in simultaneous changes in both the number of migraines and the severity of each migraine. We propose a likelihood based method applicable in diseases with recurrent events such as epilepsy or migraine headaches, where both the number of events and the severity of each event for a patient tend to depend on that patient's overall disease status. Estimation and examples are discussed.