By combining information from asthma, hay fever and eczema in the
case-control definition used in our GWAS, we expected our study design to
improve power to identify risk variants shared between, but not specific to any
of, the three diseases 6 (link). To understand if
the associations discovered in our GWAS were indeed likely to represent risk
factors shared across allergic diseases, we took advantage of the observation
that not all affected individuals report allergic co-morbidities 1 (link),22 (link),23 (link), and compared allele
frequencies between three groups of adults: asthma-only cases
(n=12,268), hay fever-only cases
(n=33,305) and eczema-only cases (n=6,276).
The studies that contributed to this analysis are indicated in Supplementary Table 1 and
described in detail in the Supplementary Note. We performed three sets of association analyses
contrasting three non-overlapping groups of individuals: asthma-only (g1) vs.
hay fever-only (g2); asthma-only (g1) vs. eczema-only (g3); and hay fever-only
(g2) vs. eczema-only (g3). These analyses are statistically independent from the
case-control analysis carried out as part of the GWAS, which facilitates
interpretation of the results. For a given sentinel SNP, results from these
analyses indicate if the risk allele is more (odds ratio [OR] >1) or less
(OR<1) common in e.g. group 1 (g1) when compared to group 2 (g2). For
example, if a SNP contributed similarly to the risks of asthma and hay fever but
not eczema, then one would expect an OR~1 in the asthma-only vs. hay
fever-only comparison, but an OR>1 in the asthma vs. eczema and hay fever vs.
eczema analyses. The significance threshold for these analyses was set at
1.2x10-4, which corresponds to a Bonferroni correction for the
136 SNPs and three sets of analyses performed (i.e.
P<0.05/(136x3)).