A study of this scope has many limitations. Here we discuss the limitations that apply to the overall risk factor analytical framework and limitations in the estimation approach for new risks and risks that have undertaken significant revisions from GBD 2015. More details and limitations of the analytical approach for each risk factor are presented in appendix 1 (p 43). First, we continue to include risk-outcome pairs that meet the World Cancer Research Fund criteria of convincing or probable evidence for causality. While these criteria have proven a useful bar for inclusion, there is an important subjective element to their interpretation. Some risk-outcome pairs included in this study might not meet these criteria or alternative criteria that are developed as new randomised trials, cohort studies, or case-control studies are published. Second, we used published cohort studies to evaluate the degree to which different risks are mediated through other risks. Estimates of pathways of mediation are used to compute the burden attributable to aggregates of risk factors such as all behavioural risks or all risks combined. While we have conducted pooled
