K-complexes, electroencephalographic markers of sleep protective mechanisms, may reflect cognitive impairment in obstructive sleep apnea (OSA). We investigated whether K-complex metrics (frequency, density, duration, amplitude) independently predict psychomotor vigilance in patients with OSA.In 140 young and middle-aged adults, K-complexes were automatically detected via a U-Net neural network analyzing F3-M2 electroencephalogram recording. K-complex metrics, including frequency (Hz), density (events/min), duration (sec) and amplitude (μV), were measured. Psychomotor vigilance was assessed via standardized Psychomotor Vigilance Test (PVT). Hierarchical regression models evaluated associations between K-complex metrics and psychomotor vigilance, adjusted for age, education, and apnea-hypopnea index (AHI).K-complex metrics showed severity-dependent associations with psychomotor vigilance in young and middle-aged adults with OSA, meaning the nature of these associations predictably changed based on OSA severity. In patients with mild/moderate OSA, lower K-complex frequency (p = 0.011) was associated with better psychomotor vigilance. In patients with severe OSA, higher K-complex density (p = 0.019) and amplitude (p = 0.030) were independently associated with better psychomotor vigilance. K-complex metrics explained 15.2 % and 23.5 % additional variance in PVT Lapses in patients with mild/moderate and severe OSA, respectively. Age, education, and AHI further modulated psychomotor vigilance, particularly in patients with mild/moderate OSA.K-complex density, frequency, and amplitude serve as severity-dependent neural biomarkers of psychomotor vigilance in young and middle-aged adults with OSA. These findings suggest the role of sleep metrics in psychomotor vigilance and highlight potential electrophysiological targets for intervention.Copyright © 2025. Published by Elsevier B.V.