Long-term oxygen therapy (LTOT) is a well established method of treatment in patients with chronic obstructive pulmonary disease (COPD). There are two main qualification criteria for LTOT: severe hypoxaemia (PaO2 < 55 mmHg) and moderate hypoxemia (PaO2 > 55 mmHg) accompanied by signs of cor pulmonale. We were unable to find any controlled investigations on the effect of LTOT on survival in this group of patients. We studied 85 COPD patients with PaO2 between 56-65 mmHg. Fourty four patients were treated at theirs homes with LTOT, 41 formed a control group. Both groups presented with severe airway obstruction--FEV1 < 1L. All patients were followed up for at least 3 years. During this time 27 patients died from the LTOT group and 26 patients from the control group. Survivors were younger and had less severe airway obstruction than nonsurvivors. Cox's proportional hazard regression analysis showed that only age and spirometric parameters--VC, VC%N, FEV1 were good predictors of survival. LTOT did not improve prognosis in this group of patients. Results of our study suggest the need of verification of LTOT qualification criteria.