To determine the effect of long term inhaled corticosteroids on lung function, exacerbations, and health status in patients with moderate to severe chronic obstructive pulmonary disease. Double blind, placebo controlled study. Eighteen UK hospitals. 751 men and women aged between 40 and 75 years with mean forced expiratory volume in one second (FEV(1)) 50% of predicted normal. Inhaled fluticasone propionate 500 microgram twice daily from a metered dose inhaler or identical placebo. Efficacy measures: rate of decline in FEV(1) after the bronchodilator and in health status, frequency of exacerbations, respiratory withdrawals. Safety measures: morning serum cortisol concentration, incidence of adverse events. There was no significant difference in the annual rate of decline in FEV(1 )(P=0.16). Mean FEV(1) after bronchodilator remained significantly higher throughout the study with fluticasone propionate compared with placebo (P<0.001). Median exacerbation rate was reduced by 25% from 1.32 a year on placebo to 0.99 a year on with fluticasone propionate (P=0.026). Health status deteriorated by 3.2 units a year on placebo and 2.0 units a year on fluticasone propionate (P=0.0043). Withdrawals because of respiratory disease not related to malignancy were higher in the placebo group (25% v 19%, P=0.034). Fluticasone propionate 500 microgram twice daily did not affect the rate of decline in FEV(1) but did produce a small increase in FEV(1). Patients on fluticasone propionate had fewer exacerbations and a slower decline in health status. These improvements in clinical outcomes support the use of this treatment in patients with moderate to severe chronic obstructive pulmonary disease.