Disrupted sleep and nocturnal hypoxia are common in cystic fibrosis (CF). However, the predictors of nocturnal hypoxia in CF are still controversial. In order to identify the risk factors for nocturnal desaturation and sleep disturbances, we carried out a clinical and polysomnographic investigation of CF patients. We studied 30 clinically stable CF cases with clinical lung disease (mean age = 12.8; mean FEV1 = 65.2), 10 CF cases without significant lung disease (mean age = 13.3; mean FEV1 = 99.8), and 20 controls (mean age = 15.5). Patients were evaluated by spirometry, 6-min walk test, the Shwachman-Kulczycki (S-K) score, and full overnight polysomnography. Cases with clinical lung disease had lower body mass index, forced vital capacity, and S-K scores. During sleep, five CF cases with clinical lung disease (15%) had SaO(2) <90% during more than 30% of total sleep time and 11 cases (36.6%) had a nadir SaO(2) below 85%. FEV1 values for CF cases with clinical lung disease were related to nadir SaO(2) (P < 0.03) and to mean SaO(2) (P = 0.02). A receiver operating characteristic (ROC) analysis determined FEV1 at 64% to be predictive of nocturnal desaturation as defined by minimum SaO(2) <85% (sensitivity = 92.3%; specificity = 77.3%) or SaO(2) <90% for 30% of sleep time (sensitivity = 81.8%; specificity = 85.2%). Frequency of impaired sleep was not different in CF cases with (N = 2) and without significant lung disease (N = 5, P = 0.53). Sleep architecture was not significantly different between the two groups. Sleep apnea was present in three CF cases with clinical lung disease and in one case without significant lung disease. In summary, desaturation during sleep can be predicted by FEV1 <64% with good sensitivity and specificity. There are no significant differences in sleep architecture between clinically stable CF cases with and without significant lung disease.