Causal Knowledge as a Prerequisite for Interrogating Bias: Reflections on Hernán et al. 20 Years Later.

This article focuses on the significance of causal knowledge in identifying and addressing bias in epidemiologic studies, particularly confounding and collider bias. It discusses the 2002 paper by Hernán et al., which emphasized using theory and directed acyclic graphs (DAGs) rather than data-driven criteria alone to guide confounder selection, highlighting that controlling for colliders—variables influenced by two other variables—can introduce bias known as collider bias. The article further explores the relationship between collider bias and selection bias, the importance of defining source populations for internal and external validity, and the need for explicit causal frameworks to distinguish different biases and improve study design. It concludes that causal knowledge is essential not only for identifying confounders but also for hypothesizing and evaluating potential sources of bias to produce valid and generalizable epidemiologic findings.