Bridging the Divide: Tackling Tensions Between Life-Course Epidemiology and Causal Inference

Abstract

Life-course epidemiologists have developed sophisticated models for how exposures throughout life—from gestation to old age—shape health, sometimes years after the exposure occurred. The field, however, has been slow to adopt robust causal inference methods, including quasi-experimental designs. This reflects, at least in part, a tension between ( a) study designs that maximize our ability to make causal claims and ( b) exposure operationalizations that correspond with life-course theories. In this narrative review, we attempt to mitigate that tension. We first discuss the unique challenges for causal inference in life-course epidemiology. We then outline how quasi-experimental methods have already contributed to testing life-course theories, as well as the limitations of the quasi-experimental methods therein. We close with solutions that bridge the gap between modern developments in causal inference and life-course epidemiology, including redefined estimands to maximize public health impact; marginal structural and structural nested models; longitudinal instrumental variables approaches; leveraging new data linkages, such as with detailed residential histories; and triangulation across methods, including adopting a pluralistic approach to causal inference.