Traditional linkage analysis and genome-wide association studies have identified HLA and a number of
non-HLA genes as genetic factors for islet autoimmunity (IA) and type 1 diabetes (T1D). However, the
relative risk associated with previously identified non-HLA genes is usually very small as measured in
cases/controls from mixed populations. Genetic associations for IA and T1D may be more accurately
assessed in prospective cohorts. In this study, 5806 subjects from the TEDDY (The Environmental Determinants of Diabetes in the Young) study, an international prospective cohort study, were genotyped
for 176,586 SNPs on the ImmunoChip. Cox proportional hazards analyses were performed to discover the
SNPs associated with the risk for IA, T1D, or both. Three regions were associated with the risk of
developing any persistent confirmed islet autoantibody: one known region near SH2B3 (HR ¼ 1.35,
p ¼ 3.58 107) with Bonferroni-corrected significance and another known region near PTPN22
(HR ¼ 1.46, p ¼ 2.17 106) and one novel region near PPIL2 (HR ¼ 2.47, p ¼ 9.64 107) with suggestive
evidence (p < 105). Two known regions (PTPN22: p¼ 2.25 106, INS; p¼ 1.32 107) and one novel
region (PXK/PDHB: p ¼ 8.99 106) were associated with the risk for multiple islet autoantibodies. First
appearing islet autoantibodies differ with respect to association. Two regions (INS: p¼ 5.67 106 and
TTC34/PRDM16: 6.45 106) were associated if the fist appearing autoantibody was IAA and one region
(RBFOX1: p ¼ 8.02 106) was associated if the first appearing autoantibody was GADA. The analysis of
T1D identified one region already known to be associated with T1D (INS: p ¼ 3.13 107) and three novel
regions (RNASET2, PLEKHA1, and PPIL2; 5.42 106 > p > 2.31 106). These results suggest that a number of low frequency variants influence the risk of developing IA and/or T1D and these variants can
be identified by large prospective cohort studies using a survival analysis approach.