For any locus identified via GWAS, we need to consider whether this reflects a potential modifiable risk factor. However, it is difficult to exclude the possibility that this locus is independently associated with both a modifiable risk factor and the disease outcome directly. For example, a recent study found an association between a polygenic risk score for schizophrenia (combining multiple variants identified with genome-wide significance into a single risk score) and cannabis use [32]. The authors concluded that this indicates that some of the association between schizophrenia and cannabis is due to a shared genetic aetiology. However, an alternative explanation could be that genetic predisposition to schizophrenia (and behaviours associated with this) increases the risk of cannabis use. Here, the distinction between mediated and biological pleiotropy is useful—the former refers to the genetic influence on the outcome operating via an exposure or intermediate phenotype, while the latter refers to a direct and independent genetic influence on both the exposure and the outcome [33]. Mediated pleiotropy is a single process leading to a cascade of downstream events, ultimately leading to a
