Oxygen (O2) has been reported to improve exercise tolerance in some patients with chronic obstructive pulmonary disease (COPD) despite only mild resting hypoxemia (PaO2 greater than 60 mm Hg). To confirm these prior studies and evaluate potential mechanisms of benefit, we measured dyspnea scores by numeric rating scale during cycle ergometry endurance testing and correlated the severity of dyspnea with right ventricular systolic pressure (RVSP) measured by Doppler echocardiography during a separate supine incremental exercise test. Both sets of exercise were performed according to a randomized double-blind crossover protocol in which patients breathed compressed air or 40% O2. We studied 12 patients with severe COPD (FEV1 0.89 +/- 0.09 L [mean +/- SEM], FEV1/FVC 37 +/- 2%, DLCO 9.8 +/- 1.5 ml/min/mm Hg[47% of predicted], PaO2 71 +/- 2.6 mm Hg). With endurance testing on compressed air, PaO2 did not change significantly in the group as whole (postexercise PaO2 63 +/- 5.1 mm Hg, p = NS), but did fall to less than 55 mm Hg in four patients from this group. Duration of exercise increased on 40% O2 from 10.3 +/- 1.6 to 14.2 +/- 1.5 min (p = 0.005), and the rise in dyspnea scores was delayed. Oxygen delayed the rise in RVSP with incremental exercise in all patients and lowered the mean RVSP at maximum exercise from 71 +/- 8 to 64 +/- 7 mm Hg (p less than 0.03). Improvement in duration of exercise correlated with decrease in dyspnea (r2 = 0.66, p = 0.001) but not with decreases in heart rate, minute ventilation, or RVSP.(ABSTRACT TRUNCATED AT 250 WORDS)