was nevertheless small, especially as fitted to a complex model like a latent growth curve model. Replication with a larger sample size is needed for more robust conclusions. However, the experimental nature of this RCT of the Fast Track project provides more statistical power than correlational studies with equal sample size, and also avoids the potential problem of gene-environment correlation (Bakermans-Kranenburg and van Ijzendoorn 2015; van Ijzendoorn and Bakermans-Kranenburg 2015). Second, only a set of tagging SNPs that provided a parsimonious summary of all NR3C1 variation was included in the analyses, as opposed to the complete sequence of the gene. Not necessarily a limitation itself because tagging aims to capture genetic variation sufficiently without extensive genotype, genome-wide data could provide the possibility of controlling for population stratification with more advanced techniques such as genetic principal components (e.g., Cleveland et al. 2015). Third, while this study only examined one gene, NR3C1, it would be more informative in future studies to investigate cumulative genetic influences using polygenic scores (e.g., Brody et al. 2013). Fourth, the current study did not examine the mechanisms underlying the NR3C1 moderation of the intervention effect. Future studies could further investigate the extent to which genetically moderated intervention
