effects. Future research should employ longitudinal designs to examine within-person changes in the role of genetic risk and social support in relation to alcohol use across development. Third, despite using a relatively large sample size compared to many other prior G×E studies, the sample size for the AA subsample may still be underpower to detect 3-way interactions for testing sex differences or developmental differences in G×E effects. Future studies are needed to replicate the present findings with larger AA samples. Furthermore, COGA is a high-risk sample of participants from extended families enriched for alcohol use disorders and findings from this study may not be generalizable to other samples with different recruitment strategies (Savage et al., 2018). Thus, our sample is not representative of the general population, which may increase the potential for collider bias in estimating the effect of alc-PRS (Akimova et al., 2021). Future research is needed to replicate our findings in community and population-based samples and to apply advanced methodologies such as principal component analysis of measured cofounders as a method to reduce collider bias (Thomas et al., 2022). Finally, we note that the effect sizes for the significant genetic main and G × E effects are quite
