OBJECTIVE To investigate associations between HLA-DQB1*0602 allele status and measures of narcolepsy symptom severity. METHODS Cross-sectional study of population-based narcolepsy patients. METHODS King County, Washington. METHODS All prevalent cases (n = 279) of physician-diagnosed narcolepsy ascertained from 2001-2005. METHODS N/A. METHODS Narcolepsy diagnosis was based on cataplexy status, diagnostic sleep study results, and chart review. The number of HLA-DQB1 alleles was determined from buccal genomic DNA. Symptom severity instruments included the Epworth Sleepiness Scale (ESS), the Ullanlinna Narcolepsy Scale (UNS), age of symptom onset, subjective sleep latency and duration, and various clinical sleep parameters. We used linear regression adjusted for African American race and an extended chi-square test of trend to assess relationships across ordered groups defined by allele number (0, 1, or 2). RESULTS Narcolepsy patients were 63% female and 82% Caucasian, with a mean age of 47.6 years (SD = 17.1). One hundred forty-one (51%) patients had no DQB1*0602 alleles; 117 (42%) had one; and 21 (7%) had two. In the complete narcolepsy sample after adjustment for African American race, we observed a linear relationship between HLA-DQB1*0602 frequency and sleepiness as defined by the ESS (P < 0.01), narcolepsy severity as defined by UNS (P < 0.001), age of symptom onset (P < 0.05), and sleep latency (P < 0.001). In univariate analyses, HLA-DQB1*0602 frequency was also associated with napping (P < 0.05) and increased car and work accidents or near accidents (both P < 0.01). Habitual sleep duration was not associated with HLA status. These race-adjusted associations remained for the ESS (P < 0.05), UNS (P < 0.01), and sleep latency (P < 0.001) when restricting to narcolepsy with cataplexy. CONCLUSIONS Narcolepsy symptom severity varies in a linear manner according to HLA-DQB1*0602 allele status. These findings support the notion that HLA-DQ is a disease-modifying gene.