Three important assumptions in Mendelian Randomization are: a) and b): G associated with X and independent of U. Thus, the effect of G on X is not affected by U. c) Given X and U, G is independent of Y. Thus, the effect of G on Y can be fully assessed by the effect of G on X and then the effect of X on Y, after adjusting for confounders U; i.e. βITT = βXGβIV as we have shown in Analysis section. (Adapted from Nitsch D, et al. ).