DIMACS Computational and Mathematical Epidemiology Seminar Series


Title: EXACT METHODS APPLIED TO GROUP SEQUENTIAL AND OTHER STRATIFIED COMPARATIVE POISSON DESIGNS

Speaker: Qi Xia, Rutgers University

Date: May 23, 2005 2:30 - 3:45 pm**

Location: DIMACS Center, CoRE Bldg, Room 431, Rutgers University, Busch Campus, Piscataway, NJ


Abstract:

Group sequential methods have been commonly applied to clinical trials with continuous or non-rare binary outcome variables, but not to comparative Poisson trials (of Poisson or rare binary outcomes) yet. We developed a series of exact group sequential methods and applied to comparative Poisson trials. The methods are based on statistical inference on the relative risk of treated vs. control groups in stratified comparative Poisson studies. We show that application of exact group sequential methods can greatly reduce the comparative Poisson trial length in terms of expected number of events needed to be observed, particularly when the null hypothesis is false. Application of curtailed early stoppage rules (with futility bounds) combined with group sequential methods further reduces the trial length particularly when the null hypothesis is true. A set of Splus programs were developed to implement the calculation of test size, power and expected number of events to be observe for exact group sequential comparative Poisson trials. Examples are given to demonstrate how to plan a group sequential trial using these Splus programs before or during a trial. These methods can be expanded to plan an exact stratified group sequential comparative Poisson trial to account for more explanatory variable(s).

Advisor DR Hoover.