Fig. 1.

Direct Acyclic Graph (DAG) outlining our causal assumptions. The green square with triangle shape represents our exposures, whereas the blue square with the rectangle represents our outcomes. Empty green squares represent ancestors of the exposure (i.e. factors we assumed to cause the exposure, but not the outcome). Based on our causal assumptions, in order to calculate the total effect of inflammatory biomarkers IL-6 and CRP on the outcome, it was necessary to adjust our models for: child’s sex, genetic risk for eating disorders, autoimmune disorders, fat mass at age nine years, mental health difficulties at age seven years, exposure to bullying at age eight as well as maternal age, highest education (as a proxy for socio-economic position), depression, and stressful life events. Grey squares represent confounders which we could adjust for, as this information is available in the ALSPAC dataset, while white circles represent variables which we were not able to observe in the data. We discuss implication of not adjusting for these variables in the limitations section. Adjustment for child’s ethnicity, in red square, was not required based on our assumptions. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)