Finally, and perhaps the biggest challenge, is the need to balance the trade-off between the need for large samples and identifying precise measures of environmental exposure. Large samples are needed to detect GxE (larger even than those needed in standard GWAS). However, large samples often lack the depth and breadth necessary to capture data on environmental or phenotype measures. Although smaller samples frequently have rich and repeated measures, they are underpowered to establish robust associations. Smaller samples can be combined to increase statistical power. However, challenges will arise in trying to harmonize measures of environment across these datasets. In other words, efforts to ensure adequate sample size for each unique combination of risk factors and GxE strata can lead to a “watered-down” environmental measure that lacks any meaningful variability; a classic example would be an instance where respondents are simply classified as “exposed” or “non-exposed.” Longitudinal birth cohort studies, which can include prospective measures of environmental exposures along with detailed phenotype data and genome-wide data, may be one promising avenue for conducting GEWIS in the future. Moreover, the growing interest
