Examining heterogeneity in treatment response in clinical trials: The poor man’s mxture model analysis
View Presentation View Presentation
*Kathleen W. Wyrwich, UnitedBioSource Corporation 

Keywords: responder definition, clinical trials, hereogeneity

In clinical trials and observational studies, post-hoc analyses provide an excellent opportunity to explore unknown or unstudied associations between the endpoints and patient characteristics measures during the trial using logistic regression. To employ this method, each endpoint needs to be dichotomized, and a reasonable responder definition or threshold needs to be selected for classifying each trial participant as improved vs. about the same or got worse on the outcome of interest. Once classified, a multiple logistic regression results can identify important patient attributes strongly associated with the outcome, as well as test interaction variables when crossed with treatment group assignment that are associated with outcomes.