The effects of 6 h of continuous low flow, nasally administered oxygen were compared with the effects of breathing air in 10 men and 2 women with obstructive sleep apnea and daytime hypersomnolence. The overall quality of sleep, sleep fragmentation, the pattern of breathing, nocturnal oxygenation, and the clinical effects on daytime hypersomnolence determined by multiple sleep latency testing were evaluated. We found that in non-REM sleep, breathing 3 L/min of oxygen increased baseline percent arterial oxyhemoglobin saturation and decreased both the rate of sleep-disordered breathing from 69 +/- 36 to 56 +/- 39 (mean +/- SD) (p less than 0.02) episodes per hour and the peak fall in arterial oxyhemoglobin saturation from 11.5 +/- 5.6% to 6.5 +/- 4.0% (p less than 0.001). In addition, oxygen significantly reduced the percentage of central and mixed sleep-disordered breathing events, thus increasing the percentage of obstructive sleep-disordered breathing events. In contrast, during REM sleep, neither the baseline nor the peak fall in oxyhemoglobin saturation during disordered breathing improved; however, there was a significant reduction in hourly sleep-disordered breeathing rate from 70 +/- 17 to 56 +/- 23 (p less than 0.02) episodes. These improvements in oxygenation and pattern of breathing were associated with improved sleep architecture characterized by a decrease in the number of awakenings from sleep and an increased total sleep time from 335 +/- 72 to 369 +/- 68 min (p less than 0.05). Although 7 of 12 patients felt more alert after oxygen therapy, there was no overall improvement in multiple sleep latency test results.(ABSTRACT TRUNCATED AT 250 WORDS)