A reliable predictor of benefit from corticosteroid treatment in patients with chronic airflow limitation is needed. In a single-blind, sequential crossover trial of placebo and prednisone (30 mg/day) treatment, with each given for 2 wk, we investigated whether an increased proportion of sputum eosinophils (>= 3%) predicts a beneficial effect of prednisone in smokers with severe obstructive bronchitis. Patients were seen before and after each treatment. Clinical measurements were made blind to the laboratory findings and vice-versa. Eighteen of 20 patients completed the study. Eight had sputum eosinophilia and similar clinical and physiologic characteristics to those of 10 patients without a finding of sputum eosinophilia. Only in patients with sputum eosinophilia did prednisone, as compared with placebo, produce a statistically significant and clinically important mean effect on effort dyspnea of 0.8 (95% confidence interval [CI]: 0.3 to 1.2), p = 0.008, and in quality of life of 1.96 (95% CI: 0.5 to 3.3), p = 0.01, associated with a small improvement in FEV1 of 0.11 L (95% CI: - 0.04 to 0.23 L), p = 0.05. In these patients, prednisone also produced a significant decline in the median sputum eosinophil percentage, from 9.7% to 0.5% (p = 0.002), eosinophil cationic protein (ECP), from 6, 000 microgram/L to 1,140 microgram/L (p < 0.001), and fibrinogen, from 25. 3 mg/L to 5.4 mg/L (p < 0.001). These findings indicate that in smokers with severe airflow limitation, sputum eosinophilia predicts a beneficial effect of prednisone treatment. Improvement in FEV1, after prednisone treatment in this population, is small, and may not be appreciated in clinical practice.