In Japanese, Koreans and Caucasians, narcolepsy/hypocretin deficiency is tightly associated with the DRB1*15:01-DQA1*01:02-DQB1*06:02 haplotype. Studies in African-Americans suggest a primary effect of DQB1*06:02, but this observation has been difficult to confirm in other populations because of high linkage disequilibrium between DRB1*15:01/3 and DQB1*06:02 in most populations. In this study, we studied human leucocyte antigen (HLA) class II in 202 Chinese narcolepsy patients (11% from South China) and found all patients to be DQB1*06:02 positive. Comparing cases with 103 unselected controls, and 110 and 79 controls selected for the presence of DQB1*06:02 and DRB1*15:01, we found that the presence of DQB1*06:02 and not DRB1*15:01 was associated with narcolepsy. In particular, Southern Chinese haplotypes such as the DRB1*15:01-DQA1*01:02-DQB1*06:01 and DRB1*15:01-DQA1*01:02-DQB1*05 were not associated with narcolepsy. As reported in Japanese, Koreans, African-Americans and Caucasians, additional protective effects of DQA1*01 (non-DQA1*01:02) and susceptibility effects of DQB1*03:01 were observed. These results illustrate the extraordinary conservation of HLA class II effects in narcolepsy across populations and show that DRB1*15:01 has no effect on narcolepsy susceptibility in the absence of DQB1*06:02. The results are also in line with a previously proposed 'HLA-DQ allelic competition model' that involves competition between non-DQA1*01:02, non-DQB1*06:02 'competent' (able to dimerize together) DQ1 alleles and the major DQα*01:02/ DQβ*06:02 narcolepsy heterodimer to reduce susceptibility.