Site Search

Colloquium - Dept. Mathematics & Statistics presents Dr. Ann Lazar of the Dept. of Epidemiology & Biostatistics, UCSF, San Francisco

Add this event into your calendar using the iCAL format
  • Thu, 12/10/2015 - 3:00pm

Title:  Identifying heterogeneity of treatment effects in the era of precision health

Abstract:  Patients and clinicians often make treatment decisions based on data representing the  average (or overall) study results. This decision assumes that the average study population represents an individual patient even though some patients may differ from the overall study results, in terms of demographic or health behavior characteristics, known as heterogeneity of treatment effects or HTEs. Certainly identifying HTEs is central to helping patients make informed personalized decisions about their health. Guidelines for traditonal heterogeneity evaluation recommend introducing an interaction term between treatment and dichotomized patient characteristic in a regression model. However, this traditional HTE method may overlook complex interaction effects and distill a complex interaction effect to a p-value of a regression parameter. Alternative HTE approaches can be used to graphically illustrate complex patterns of treatment  effect heterogeneity--Subpopulation Treatment Effect Pattern Plots (STEPP) and Johnson-Neyman. An advantage of STEPP is that it is a non-parametric alternative to the Johnson-Neyman approach that can now be used to focus on disease specific events in the competing risks setting based on observed minus expected methodology. Examples from randomized clinical trials (RCTs) will be provided to illustrate how STEPP and the Johnson-Neyman methodology can be used to explore graphically the pattern of heterogeneity as a function of a patient characteristic measured on a continuous scale.

Dina Labelle
519-253-3000 ext.3016