The International Classification of Sleep Disorders (ICSD-2) criteria for low CSF hypocretin-1 levels (CSF hcrt-1) still need validation as a diagnostic tool for narcolepsy in different populations because inter-assay variability and different definitions of hypocretin deficiency complicate direct comparisons of study results. Interviews, polysomnography, multiple sleep latency test, HLA-typing, and CSF hcrt-1 measurements in Danish patients with narcolepsy with cataplexy (NC) and narcolepsy without cataplexy (NwC), CSF hcrt-1 measurements in other hypersomnias, neurological and normal controls. Comparisons of hypocretin deficiency and frequency of HLA-DQB1*0602-positivity in the Danish and eligible NC and NwC populations (included via MEDLINE search), by (re)calculation of study results using the ICSD-2 criterion for low CSF hcrt-1 (< 30% of normal mean). In Danes, low CSF hcrt-1 was present in 40/46 NC, 3/14 NwC and 0/106 controls (P < 0.0001). Thirty-nine of 41 NC and 4/13 NwC patients were HLA-DQB1*0602-positive (P < 0.01). Hypocretin-deficient NC patients had higher frequency of cataplexy, shorter mean sleep latency, more sleep onset REM periods (P < 0.05) and more awakenings (NS) than did NC patients with normal CSF hcrt-1. Across populations, low CSF hcrt-1 and HLA-DQB1*0602-positivity characterized the majority of NC (80% to 100%, P = 0.53; 80% to 100%, P = 0.11) but a minority of NwC patients (11% to 29%, P = 0.75; 29% to 89%, P = 0.043). The study provides evidence that hypocretin deficiency causes a more severe NC phenotype. The ICSD-2 criterion for low CSF hcrt-1 (< 30% of normal mean) is valid for diagnosing NC, but not NwC. HLA-typing should precede CSF hcrt-1 measurements because hypocretin deficiency is rare in HLA-DQB1*0602-negative patients.