The approaches described here can also be informative with respect to null results. If a modifiable exposure is under genetic influence and is also causally related to a disease outcome, we would expect to eventually see genetic variants associated with the exposure emerge in a GWAS of the outcome, given sufficient sample size. If this is not seen, this suggests that there may be no causal pathway operating (or that any causal relationship is very weak). Of course, interpreting null results must be done cautiously, particularly in cases where the prevalence of the modifiable exposure or the minor allele frequency differs across populations. Current GWAS cannot control for these sources of heterogeneity, which may impact the power of GWAS to identify modifiable exposures in the way we have described. Cross-contextual comparisons (e.g., across GWAS conducted in different populations) may be informative in these cases.
