Causal inference in Developmental Origins of Health and Disease (DOHaD) research

Abstract

Studies of the developmental origins of health and disease (DOHaD) often rely on prospective observational data, from which associations between developmental exposures and outcomes in later life can be identified. Typically, conventional statistical methods are used in an attempt to mitigate problems inherent in observational data, such as confounding and reverse causality, but these have serious limitations. In this review, we discuss a variety of methods that are increasingly being used in observational epidemiological studies to help strengthen causal inference. These methods include negative controls, cross-contextual designs, instrumental variables (including Mendelian randomization), family-based studies, and natural experiments. Applications within the DOHaD framework, and in relation to behavioral, psychiatric, and psychological domains, are considered, and the considerable potential for expanding the use of these methods is outlined.