Importantly, before assessment of clinical validity is made, the contribution of genetic sum scores, rather than individual associated SNPs, must be determined. The finding that genetic sum scores created from SNPs meeting less stringent p-value thresholds were significantly associated with AD and had significant discriminative ability suggests that varying p-value thresholds could better detect variants of small effect. However, it is difficult to distinguish true alleles of vanishingly small effect from alleles in LD with causal alleles. Because variants contributing to AD have small effect sizes, and the outcome used in the association studies is a dichotomous diagnosis rather than a continuous outcome, larger sample sizes are needed for increased power to detect causal variants that replicate across studies (Bierut et al., 2010). The samples used in this study did not have enough power to detect the entire range of small effect sizes for individual variants assessed in these analyses at a genome-wide significance level. Splitting the COGA-SAGE combined sample further reduced power.
