CMStatistics 2017: Start Registration
View Submission - CMStatistics
B0171
Title: Time-to-event mediation analysis with repeatedly measured mediators Authors:  Stijn Vansteelandt - Ghent University and London School of Hygiene and Tropical Medicine (Belgium) [presenting]
Abstract: The LEADER trial found protective cardiovascular effects of liraglutide compared to placebo in patients with Type II diabetes and high cardiovascular risk. Effects were also found on glycated hemoglobin, body weight, blood pressure and heart rate, thereby raising the question to what extent these potential pathways my explain liraglutide's protective effect. We will explain how we addressed this question by expanding modern techniques from causal mediation analysis. In particular, we will show how to identify and infer the natural indirect of liraglutide on the time to major cardiovascular events via the repeatedly measured glycated hemoglobin levels. The considered proposal addresses complications due to patients dying before the mediator is assessed, due to the mediator being repeatedly measured, and due to post-treatment confounding of the effect of glycated hemoglobin by other mediators, which makes mediation analysis a challenging enterprise.