A long-term prognostic study of 129 patients with severe chronic obstructive pulmonary disease (initial value for forced expiratory volume in 1 sec [FEV1] smaller than or equal to 1,000 ml) is reported. Data from the patients (72 per patient) were obtained in a clinically stable phase during the first hospital admission and subsequent regular outpatient visits; the data were processed using multiple regression and discriminant analysis. Five- and 10-year cumulative survival rates were 69 and 40 per cent, respectively. Prediction of survival rate was determined mainly by the rate of decrease in initial FEV1 per year and the degree of increase of initial FEV1 after administration of thiazinamium, a bronchodilator. Initial FEV1 values per se contributed to prediction only when they were less than 450 ml. Study of the data from the individual patients showed the presence of different types of decrease in FEV1: linear, exponential, combined linear-exponential, and phasic. FEV1 sometimes remained unchanged for several years or even showed an increase. No sudden decreases were observed. The better survival rate in patients with greater reversibility of airway obstruction after administration of thiazinamium has important bearing on treatment policy. The hypothesis is put forward that the better survival rate in the present study might be the result of continuing medical care. Comprehensive treatment, provided it includes a component directed at preventing and treating reversible airway obstruction, may be of major importance in determining survival in patients with severe chronic obstructive pulmonary disease. The value of so-called dominant, survival-predicting factors is considered to be of secondary importance; their main function is to alert the physician to the fact that prognosis as concerns survival may be unfavorable and that preventive and therapeutic measures should be taken without delay.