Causal Mediation Analysis: A Rigorous Framework for Decomposing Exposure Effects in Dementia Epidemiology (Motivated by the HUNT Study on Parity and Dementia Risk by Mekonnen et al.)

Causal mediation analysis provides a rigorous framework for dissecting the pathways through which an exposure exerts its effects on an outcome—an especially critical endeavor in observational epidemiology where complex, multifactorial relationships prevail. This report introduces the conceptual foundations and methodological execution of causal mediation analysis, using a population-based study on parity and dementia risk from the HUNT cohort in Norway as a motivating example. The study employed inverse odds weighting to evaluate whether midlife psychosocial, socioeconomic, lifestyle, and health-related factors mediated the observed U-shaped association between the number of children and late-life dementia risk. Despite robust longitudinal data and comprehensive adjustment for confounding, the mediation analysis revealed that these potential pathways did not explain the elevated dementia risk among childless individuals or those with one or four or more children. This suggests that the relationship may be primarily direct or influenced by unmeasured mechanisms. The report highlights key methodological requirements—including proper temporal sequencing, confounder control, and sensitivity analyses—and provides visual and tabular summaries of effect decomposition. It concludes by outlining limitations and future directions, emphasizing the importance of enhancing methodological flexibility to address time-varying mediators and complex causal structures in life-course epidemiologic research.