The association of narcolepsy with HLA-DQB1*0602 is established in Japanese, African-Americans, European, and North American Caucasians. We examined DRB1, DRB3, DRB4, DRB5, DQA1, and DQB1 in 163 patients with centrally mediated daytime sleepiness (100 with narcolepsy) and 211 Korean controls. In this population, the DQB1*0602 association was always evident in the context of the DRB1*1501-DQA1*0102-DQB1*0602 haplotype. The DQB1*0602 association was highest in cases with hypocretin deficiency (100% vs 13% in controls), most of which had narcolepsy-cataplexy (81%). A weaker DQB1*0602 (45%) association was present in cases without cataplexy. No human leukocyte antigen (HLA) association was present in idiopathic hypersomnia or in cases with normal cerebrospinal fluid (CSF) hypocretin-1. As in other populations, DQB1*0602 homozygosity increased risk in cases with cataplexy and/or hypocretin deficiency (odds ratio = 2.0 vs heterozygotes). Non-DQB1*0602 allelic effects were also observed but could not be interpreted in the context of DQB1*0602 overabundance and linkage disequilibrium. We therefore next analyzed compound heterozygote effects in 77 subjects with either hypocretin deficiency or cataplexy and one copy of DRB1*1501-DQA1*0102-DQB1*0602, a sample constructed to maximize etiologic homogeneity. In this analysis, we found additional predisposing effects of DQB1*0301 and protective effects for DQA1*0103-DQB1*0601. Unexpectedly, the predisposing effects of DQB1*0301 were present in the context of various DQA1-bearing haplotypes. A predisposing effect of DQA1*0303 was also suggested. These results indicate a remarkable consistency in the complex HLA association present in narcolepsy across multiple ethnic groups.