Caution is also required in rejecting the role of a SNP based on a failed replication attempt (91). In our case, the use of the relatively homogeneous founder population might have facilitated the detection of associations, but some of the identified SNPs may be particularly difficult to replicate in more heterogeneous populations. For example, the SNPs we identified might be in LD with the functional variant in the Sardinia but not in other populations. Some of the associations we identified might also depend on population-specific genetic or environmental background (gene-gene and gene-environment interactions). The replication samples also differed in recruitment strategies and used shorter and slightly less reliable phenotypic measures, which might have reduced power and contributed to differences across samples. In fact, heterogeneity among candidate gene studies has been linked to differences in the instrument used (92), but when we scored the SardiNIA data for the shorter questionnaires versions we found little differences in the results. Furthermore, given the universality of the five factors (17) and the commonality of genetic factors for each of a variety of other complex
