Although we do demonstrate a significant case-control association between rs279871 and CD within our Caucasian/Hispanic sample, the result does not survive correction for multiple comparisons (i.e. Bonferonni correction for only 3 tests α = 0.05/3 = 0.017), is in the opposite direction than the Dick et al., 2006 finding (association of CD with GG genotype) and is not supported by our family-based tests for association. Each approach, case-control and family-based, has strengths and weaknesses. Our case-control analyses, compared with family-based, provide greater power to detect association under a dominant model with an odds ratio of 2.0 (Dick et al., 2006) but population substructure may potentially lead to false positive results. Alternatively, the family-based approach protects against the confound of population stratification but is less powerful. Our power calculations for our family-based analyses suggest that we had adequate power to detect an association under some but not all models/assumptions; for an odds ratio of 2.0 and assuming a dominant model (Dick et al., 2006), our family-based association tests had only 66% power.
