Thomas GerdsUniversité de Copenhague
Le 11 sept. 2025
Titre :
Longitudinal Causal Inference in Pharmacoepidemiology
Résumé :
A central aim of pharmacoepidemiology is to evaluate the effectiveness
and safety of medications using electronic health records. For example,
studies of glucose-lowering medications in patients with type 2 diabetes
can be designed to emulate a target trial comparing different drug
classes, with cardiovascular disease events as the primary
outcome. Compared with clinical trials, pharmacoepidemiologic studies
benefit from the diversity and size of the studied populations and the
possibility to examine long-term outcomes. At the same time, careful
application of the emulated target trial framework is needed to
interpret results and to recognize the inherent limitations of such data
where people start, switch, pause and stop treatment. Equally important
for causal inference is the discretization of complex longitudinal data
in a way that preserves essential information while avoiding causal
confusion.
