We may draw on the analogy with MR-Egger to contemplate other approaches based on the ratio of prognosis effects to incidence effects. Such approaches would entail other assumptions that require careful consideration. For example, a counterpart of the median ratio estimator34 would assume that at least half of the SNPs considered have no direct effect on prognosis. Alternative approaches related to parallel work in Mendelian randomisation are a promising area for further development.
