The evolution of selection bias in the recent epidemiologic literature-a selective overview

Abstract

Selection bias has long been central in methodological discussions across epidemiology and other fields. In epidemiology, the concept of selection bias has been continually evolving over time. In this issue of American Journal of Epidemiology, Mathur and Shpitser (Am J Epidemiol. 2025;194(1):267-277) present simple graphical rules for assessing the presence of selection bias when estimating causal effects by using a single-world intervention graph (SWIG). Their work is particularly insightful as it addresses the scenarios where treatment affects sample selection-a topic that has been underexplored in previous literature on selection bias. To contextualize the work by Mathur and Shpitser, we trace the evolution of the concept of selection bias in epidemiology, focusing primarily on the developments in the last 20-30 years following the adoption of causal directed acyclic graphs (DAGs) in epidemiologic research.

Keywords: causal directed acyclic graph; causal inference; collider bias; epidemiologic research; selection bias; single-world intervention graph.