Night-to-night variability of apneas on overnight polymnography exists in patients with documented obstructive sleep apnea (OSA). In this study, we evaluated the possibility that this variability may be severe enough to miss the diagnosis of OSA in patients clinically at risk for the disease. We prospectively studied 11 patients who were deemed on clinical grounds to have probable OSA, but had a negative result on overnight polysomnography. Six of the 11 patients were found to have a positive second study with a significant rise in the apnea/hypopnea index (AHI) from 3.1 +/- 1.0 to 19.8 +/- 4.7 (mean +/- SEM, p < 0.01). The cause of the negative first study in these patients is unclear, but it does not seem related to risk factor pattern, sleep architecture, or test interval. The change in AHI was not found to be rapid eye movement (REM)-dependent. This study demonstrates that a negative first-night study is insufficient to exclude OSA in patients with one or more clinical markers of the disease.