The rapid growth in genome-wide association studies (GWAS) has resulted in the identification of common genetic variants associated with behavioural traits, from biomarker phenotypes that capture the downstream consequences of behaviour (e.g., body mass index [BMI]) [1] to the behaviours themselves (e.g., tobacco and alcohol use) [2]. While the success of GWAS has generated insights into the biological mechanisms underpinning these traits (see Box 1), it is less appreciated that it has also begun to tell us about the causal effects of modifiable or environmental influences on these traits. For example, a genetic variant at a locus containing the NPC1L1 gene is strongly associated with low-density lipoprotein (LDL) cholesterol level as well as with the risk of cardiovascular events. This is not because NPC1L1 is independently associated with cardiovascular problems, but simply because high cholesterol is a causal risk factor for the disease [3–5]. In other words, there are a number of cases where GWAS of disease outcomes have identified loci that capture modifiable risk factors rather than direct biological pathways. Here, we explain how this insight can inform the interpretation of GWAS results.
