Among patients with similar degrees of obstructive sleep apnea (OSA) there is considerable variability in the degree of associated nocturnal hypoxemia. The factors responsible for this variability have not been clearly defined. Therefore we studied 44 patients with OSA to identify the physiological determinants of nocturnal arterial O2 saturation (SaO2). All patients underwent pulmonary function testing, arterial blood gas analysis, and overnight polysomnography. Mean nocturnal SaO2 ranged from 96 to 66% and apnea-hypopnea index from 11 to 128 per hour of sleep. Several anthropometric, respiratory physiological, and polysomnographic variables that could be expected to influence nocturnal SaO2 were entered into a stepwise multiple linear regression analysis, with mean nocturnal SaO2 as the dependent variable. Three variables [awake supine arterial PO2 (PaO2), expiratory reserve volume, and percentage of sleep time spent in apnea] were found to correlate strongly with mean nocturnal SaO2 (multiple R, 0.854; P less than 0.0001) and accounted for 73% of its variability among patients. Body weight, other lung volumes, and airflow rates influenced awake PaO2 and expiratory reserve volume but had no independent influence on nocturnal SaO2. In a further group of 15 patients with OSA a high correlation was obtained between measured nocturnal SaO2 and that predicted by the model (r = 0.87; P less than 0.001). We conclude that derangements of pulmonary mechanics and awake PaO2 (generally attributable to obesity and diffuse airway obstruction) are of major importance in establishing the severity of nocturnal hypoxemia in patients with OSA.