To determine whether studies of breathing and oxygenation during sleep are clinically useful, we have assessed whether the detection of excess nocturnal hypoxemia in patients with chronic obstructive pulmonary disease (COPD) is of prognostic importance. Ninety-seven patients with COPD were followed for 32 to 108 (median, 70) months after studies of overnight oxygenation. Significant relationships (p less than 0.001) were obtained between mean oxygen saturation (SaO2) asleep and awake. There was similarly a significant relationship between lowest SaO2 asleep and awake, but this relationship was improved by the inclusion of awake arterial carbon dioxide tension (PaCO2). The patients who were more hypoxic at night than predicted from these regression relationships had similar survivals to the patients who were less hypoxic at night than predicted, whether excess nocturnal hypoxia was defined in terms of mean or lowest SaO2 during sleep. In the 66 patients who did not subsequently receive long-term oxygen therapy, none of the indices of nocturnal oxygenation was related to survival, the only significant predictor of survival being daytime arterial oxygen tension (PaO2). For all 97 patients, both mean nocturnal SaO2 and lowest SaO2 during sleep were related to survival (p less than 0.05), and percent predicted vital capacity was also related to survival (p less than 0.05). Neither of the oxygen saturations during sleep significantly added to the more readily and cheaply measured percent predicted vital capacity in determining survival in these patients. Thus, in patients with COPD, excess nocturnal hypoxemia is not associated with an impaired prognosis, and so studies of oxygenation during sleep cannot be recommended in the routine clinical management of these patients.