We hypothesized that regular use of long-acting beta-agonists could delay recognition of ("mask") increasing airway inflammation. We studied steroid-sparing and "masking" effects of salmeterol versus placebo in 13 asthmatic individuals requiring >= 1,500 microgram inhaled corticosteroid daily. Corticosteroid doses were reduced weekly until criteria were met for an exacerbation or the corticosteroid was fully withdrawn. Subjects were restabilized on their original dose of inhaled corticosteroid for 4 wk before crossover to the alternative treatment. Subjects maintained symptom and peak expiratory flow (PEF) diaries, and underwent weekly spirometric, methacholine challenge, sputum eosinophil, and serum eosinophil cationic protein (ECP) measurements. Mean corticosteroid dose was reduced by 87% during salmeterol treatment, versus 69% with placebo (p = 0.04). Sputum eosinophils increased before exacerbation despite stable symptoms, FEV1, and PEF. In the week before clinical exacerbation, sputum eosinophil counts were higher in the salmeterol-treatment arm (19.9 +/- 29.8% [mean +/- SD], versus placebo 9.3 +/- 17.6%; p = 0.006). Five subjects showed > 10% sputum eosinophilia before exacerbation during salmeterol treatment, as compared with two receiving placebo. In this model, salmeterol controlled symptoms and lung function until inflammation became significantly more advanced. We conclude that the bronchodilating and symptom-relieving effects of salmeterol can mask increasing inflammation and delay awareness of worsening asthma.