Conditional analysis, conditioning on the SNP with the strongest evidence of association, and adding additional SNPs into the model.

For each JIA subtype, the most significant SNP from single point analysis (Table 1) was used to start the conditional analysis, and then the most significant SNP achieving at least suggestive significance (p<0.0189) was added into the conditional model until no further SNPs were identified. In parentheses are reported the empirical p-values obtained by running 100,000 permutations.