In this study, association analyses with FTND score were conducted with the UNPHASED program (version 3.12) [Dudbridge, [2008]], which uses a log likelihood retrospective regression model. Since the VAANX-ND sample contained individuals selected for high IANGF scores, we thought it prudent to control for its effect by including the IANGF score as a covariate in the regression model. In all association analyses, including those performed for SymAlcAD and SymCanAD, a standard Z-score transformation of raw scores was performed for the phenotypes. For association tests, the SNP spectral decomposition (SNPSpD) method [Nyholt, [2004]] was used to evaluate the effective number of independent tests and P-values reported were uncorrected. Multi-marker haplotype analyses were conducted similarly with the UNPHASED program, and 1% was used as the cutoff for minor haplotypes. Pairwise LD was estimated for all subjects by the HaploView 4.0 software [Barrett et al., [2005]]. Hardy-Weinberg equilibrium tests were also conducted with the HaploView program. To exclude the impact of FTND on the association of SymAlcAD and SymCanAD, FTND scores were included as a covariate when association with SymAlcAD and SymCanAD was analyzed.
