To determine the value of theophylline in the maintenance therapy of patients with severe chronic obstructive pulmonary disease (COPD), we conducted a trial of theophylline therapy withdrawal in 38 clinically stable patients with severe COPD (FEV1 < 60 percent) predicted. Symptoms, lung function, blood gases, and 6-min walking distance were assessed on days 1 and 2. Quality of life and overall dyspnea were scored using four different clinical rating scales. Theophylline therapy was continued in 20 patients and replaced by placebo from day 3 on in the remainder; measurements were repeated on days 5 and 6. Withdrawal of theophylline therapy resulted in significant (p < 0.05) deterioration in lung function, exercise performance, and two indices of overall dyspnea, and a significant increase in scoring of symptoms and auscultation findings. Individual analysis revealed a clinically relevant deterioration in 72 percent of patients from whom theophylline therapy was withdrawn, while only 15 percent of patients receiving theophylline exhibited deterioration. No major side effects were observed. Our data show that about half of the patients with severe COPD can be considered as theophylline responders. The response of these patients to withdrawal of theophylline therapy suggests that the clinical effectiveness of this drug cannot be explained exclusively by bronchodilation. Due to the inherent difficulties in predicting response to theophylline, its effectiveness in patients with severe COPD should be determined individually, including assessment of exercise performance and ratings of dyspnea.