In a Mendelian randomization study, association 1 and 2 are estimated to determine the causal effect of the exposure (ie, the intermediate factor) on the outcome. Association 3 represents the usual association estimated in traditional observational studies which may be prone to the usual biases (ie, confounding, reverse-causality).
