The transmission disequilibrium test was used to analyze haplotypes for association and linkage to diabetes within families from the Human Biological Data Interchange type 1 diabetes repository (n = 1371 subjects) and from the Norwegian Type 1 Diabetes Simplex Families study (n = 2441 subjects). DQA1*0102-DQB1*0602 was transmitted to 2 of 313 (0.6%) affected offspring (P < 0.001, vs. the expected 50% transmission). Protection was associated with the DQ alleles rather than DRB1*1501 in linkage disequilibrium with DQA1*0102-DQB1*0602: rare DRB1*1501 haplotypes without DQA1*0102-DQB1*0602 were transmitted to 5 of 11 affected offspring, whereas DQA1*0102-DQB1*0602 was transmitted to 2 of 313 affected offspring (P < 0.0001). Rare DQA1*0102-DQB1*0602 haplotypes without DRB1*1501 were never transmitted to affected offspring (n = 6). The DQA1*0101-DQB1*0503 haplotype was transmitted to 2 of 42 (4.8%) affected offspring (P < 0.001, vs. 50% expected transmission). Although DRB1*1401 is in linkage disequilibrium with DQB1*0503, neither of the two affected children who carried DQA1*0101-DQB1*0503 had DRB1*1401. However, all 13 nonaffected children who inherited DQA1*0101-DQB1*0503 had DRB1*1401. In a case-control comparison of patients from the Barbara Davis Center, DQA1*0101-DQB1*0503 was found in 5 of 110 (4.5%) controls compared with 3 of 728 (0.4%) patients (P < 0.005). Of the three patients with DQB1*0503, only one had DRB1*1401. Our data suggest that both DR and DQ molecules (the DRB1*1401 and DQA1*0102-DQB1*0602 alleles) can provide protection from type 1A diabetes.