We retrospectively analyzed night-to-night variability in the indices of sleep apnea in a group of men who underwent consecutive polysomnograms (PSGs) in the evaluation of impotence. The study group consisted of 37 subjects. Fifty-seven percent of the subjects had an apnea/hypopnea index (AHI) of 5 or more on the first PSG, whereas 70% met this criterion on the second study. On both PSGs, 49% of the subjects exhibited an AHI of 10 or more. The AHI varied by 10 or more between the two PSGs in 32% of the cases. Using a threshold AHI of 5 or more to establish a diagnosis of sleep apnea, 22% of the subjects would not have been diagnosed by the first PSG, and the negative rate for the first PSG was 50%. The variability observed in the AHI could not be explained by differences in total sleep time; sleep stages [1 through 4 and rapid eye movement (REM) sleep]; the amount of time sleeping supine, or a combination of sleep stage and position. The mean AHI, the apnea/hypopnea-related nadir, and the mean oxyhemoglobin saturation did not differ among the two PSGs. Our observations support the notion that for groups of subjects the mean AHI is relatively constant across 2 nights of study in the sleep laboratory. However, when an AHI of 5 or 10 or more is the sole criterion used to establish the diagnosis of sleep apnea, a single PSG may not be sufficient to rule out the presence of a sleep apnea syndrome.