Withdrawal of inhaled corticosteroids is known to worsen disease control in bronchial asthma but similar data are lacking in chronic obstructive pulmonary disease (COPD). We hypothesized that clinical exacerbations requiring treatment would occur more often in patients whose inhaled corticosteroids were stopped than in other patients not treated with these agents. We studied 272 patients in mean age 65 (SD 0.8) years, mean FEV1 42.8 (SD 12.6)% predicted, entering the run-in phase of the Inhaled Steroids in Obstructive Lung Disease (ISOLDE) trial. All had been clinically stable for at least 6 weeks and there were no differences in the degree of bronchodilator reversibility, baseline lung function or pack-years of smoking between the 160 patients receiving inhaled corticosteroids and those not so treated. Inhaled corticosteroids were withdrawn in the first week of the study and during the remaining 7 weeks of the study 38% of those previously treated with these drugs had an exacerbation compared to 6% of the chronically untreated group. Patients receiving inhaled corticosteroids reported a longer duration of symptoms but neither this or any other recorded variable predicted the risk of exacerbation. These data suggest that abrupt withdrawal of inhaled corticosteroids should be monitored carefully even in patients with apparently irreversible COPD.