“…Existing mediation methods are able to quantify the single-mediation effects on a dichotomous outcome (Imai et al, 2010;Valeri & Vanderweele, 2013;Vanderweele & Vansteelandt, 2010), but may require a rare outcome in order to estimate the effects (Valeri & Vanderweele, 2013;Vanderweele & Vansteelandt, 2010). When multiple mediators are considered, the existing mediation methods may be computationally intensive or may require additional sensitivity parameters to estimate the relationship between the two mediators (Daniel et al, 2015;Wang et al, 2012). The proposed method differs from other methods in that it can estimate the probability of a dichotomous outcome, and therefore, any effect involving a contrast of the probability under two exposure status, such as the risk difference or risk ratio, without the rare outcome assumption and additional assumptions regarding the relationship between the mediators.…”