We examined the effects of arousal- and desaturation-based scoring criteria on the apnea-hypopnea index (AHI) and on the measured prevalence of obstructive sleep apnea (OSA). Ninety-four randomly selected patients underwent overnight polysomnography. Studies were scored according to three different criteria for hypopnea, as defined by a >= 10 s discernible reduction in thoracoabdominal movement associated with: (1) >= 4% decrease in oxygen saturation (SaO2) (Type A); (2) either a >= 4% decrease in SaO2 or an arousal (Type B); or (3) electroencephalographically based arousal alone (Type C). Excellent correlation existed between AHI-A, AHI-B, and the oxygen desaturation index (ODI) (r > 0.98). AHI-A and AHI-B differed by only 2.04 +/- 1.72/h (2 SD). AHI-A and AHI-B differed from the ODI by 1.04 +/- 4.07/h and 3.07 +/- 4.30/h, respectively. Despite these small differences, use of the Type B rather than Type A definition resulted in an extra case of OSA being diagnosed for every 14 to 31 patients tested, depending on the definition of OSA (AHI: >= 5, 10, 15, or 20/h). The addition of arousal-based scoring criteria for hypopnea causes only small changes in the AHI, but if OSA is defined solely by an AHI value, the measured prevalence of OSA will increase.