Chunk #46 — Reasons to be Concerned about the Published cGxE Literature — Problems with the Recipe: Statistical Concerns in cGxE Research — The importance of covariates
Failure to properly control for potential confounds can also be problematic in cGxE research. In non-experimental research, researchers typically enter potential confounding variables (e.g., gender, ethnicity, socioeconomic status, genotype quality, etc.) into regression equations to control for their effects. However, this approach controls only for the additive effects of covariates; it does nothing to control for the potential confounding effects these covariates might have on the interaction itself (Keller, 2014; Yzerbyt, Muller, & Judd, 2004). To properly control for confounders in cGxE research, investigators must also evaluate all relevant gene-by-covariate and environment-by-covariate interaction terms. To date, virtually no cGxE studies have appropriately controlled for all covariate interactions (Keller, 2014). This failure to include covariates is particularly concerning in mixed-ethnicity samples, where stratification can not only produce spurious genetic main effect association to be detected (Price, Zaitlen, Reich, & Patterson, 2010), but can also cause ethnicity-by-environment interactions to appear as spurious gene-by-environment interactions. This is because the frequency of alleles naturally varies across ethnic populations and, in the presence of a coincidental excess of affected individuals belonging to one ethnic group, spurious associations and interactions with polymorphisms of no functional consequence, except a degree of natural ethnic variation, may emerge.
