We tried to use eosinophil counts in induced sputum samples as a marker of airway inflammation, and as a guide for reducing inhaled corticosteroids in patients with well-controlled persistent asthma. The eosinophil count in induced sputum smears was defined as follows: Eos%; eosinophil percentage of 200-400 leukocytes in properly cell-separated fields, TEC; total eosinophil counts in the 5 most eosinophil-dense high power view fields (x 400). First, the eosinophil count in induced sputum samples was compared between 29 asthmatic subjects treated with inhaled corticosteroid and 15 age- and sex-matched healthy controls. Second, inhaled corticosteroid was reduced by 50% in 20 patients with green-zone asthma (morning PEF > 80% of patient's best PEF). PEF measurements were followed prospectively for 12 weeks thereafter. Once PEF decreased below 70% of their best PEF, subjects were considered as treatment "failures". Both Eos% and TEC were significantly higher than in the controls, even in well-controlled (morning PEF > 80% of their best) asthmatic patients (p = 0.001, 0.03). The chance of treatment "failure" was significantly higher in those having more eosinophils (Eos% > 10%, TEC > 100) in their initial induced sputum sample (p = 0.03, 0.001). Airway inflammation still persists in many well-controlled chronic asthmatic patients, and induced sputum eosinophilia predicts an early decrease of PEF after reduction of inhaled corticosteroids.