The critical assumption of our approach is that direct genetic effects on prognosis are independent of those on incidence. Since GWAS of susceptibility have been motivated by the discovery of novel treatment targets, our assumption may seem incompatible with the premise of GWAS. Indeed, shared pathways of incidence and prognosis have been observed in coronary heart disease, in which statins have proved effective in preventing both initial and recurrent events31. For phenotypes related to cumulative effects of long-term exposures, such as CVD but also perhaps some psychiatric traits, such shared pathways may be common. But for conditions in which prognosis depends upon the response to an initiating event, as perhaps in cancer or infectious diseases, the determinants of prognosis are conceivably independent of those for incidence. Even in CVD, determinants of arterial plaque development may be independent of those for plaque rupture. For immune-mediated disease, where a break in immunological tolerance is the key event at disease initiation, it is expected that other pathways drive disease course, since tolerance can only be broken once to a particular antigen. Also, where
