The purpose of this study was to assess the ability of continuous nocturnal oximetry to detect sleep apnea syndrome (SAS) and to recognize nonapneic oxyhemoglobin desaturations. Oxygen saturation oscillations, related to successive apneas in SAS or to apneic episodes in COPD or restrictive patients, were quantified using a new index: delta = 1/n sigma 1 n magnitude of delta(SaO2)/delta(t)(12-s intervals) Twenty-six patients (15 SAS, 8 COPD, and 3 restrictive patients) were included in a prospective study comparing nocturnal oximetry and polysomnography over 34 nights. In apneic patients, we found a strong correlation (r2 = 0.73, p less than 0.01) between time spent in apnea and the delta index. In COPD, the number of apneas was also correlated to the delta index (r2 = 0.92, p less than 0.01). A lower threshold for delta of 1.5 is accurate enough to detect apneas if initial SaO2 is greater than 93 percent. If initial SaO2 is greater than 93 percent, the delta threshold should be 0.8 (sensitivity 95 percent). Such a method could contribute to the accurate selection of patients for polysomnography.