CFE 2019: Start Registration
View Submission - CMStatistics
B0302
Title: Studying the influence of time-varying treatment regimes on a time-to-event outcome using Danish registry data Authors:  Sarah Friedrich - University Medical Center Goettingen (Germany) [presenting]
Abstract: In a Danish nationwide study in diabetes patients based on registry data, the aim was to compare the effect of different treatment regimes on a subsequent time-to-event outcome. However, definition of treatment regimes is complicated due to patients switching back and forth between treatments. Moreover, the competing risk of death further complicates analyses. We employ a nested case-control design which allows for the definition of comparable treatment histories for cases and controls. Gaining a causal interpretation of the effect estimates is not straightforward in this setting due to the complicated treatment histories, which require advanced methods of causal inference (like targeted learning) in a time-continuous framework. We discuss the strengths and weaknesses of the approach and give an outlook on the methods and developments needed to gain a causal interpretation in such a setting.