Linear regression analysis of our a priori hypothesis to examine 17 SNPs in the CHRNB3 gene revealed significant associations between several SNPs and the “dizziness” measure, adjusted for age, sex, and case status (Table 2). Five SNPs (rs10958726, rs1955186, rs1955185, rs13277254, and rs13277524) were nominally significant at p<0.01, and are likely to represent the same signal since they are in high linkage disequilibrium (r2 >0.98). Four other SNPs in the same haplotype block (rs5005909, rs6474412, rs6474413, and rs4950) were significant at the nominal p<0.03, with the minor differences in p-values likely due to slight differences in missing data (see Supplementary Figure 1). The other two SNPs in the haplotype block (rs1530847 and rs13280301) were slightly less common than the others, with no evidence of association (p>0.15). PLINK was used for set-based permutation testing to correct for all 17 SNPs examined in this gene (Purcell et al., 2007). rs1955186 emerged as the SNP with the lowest p value, with a gene-based empirical p=0.021 (Table 3).
