To determine predictors for failed reduction of inhaled corticosteroids (ICS), in 50 subjects with well-controlled asthma (age 43.7 [18-69]; 22 males) taking a median dose of 1,000 microg ICS/d (100-3,600 microg/d), ICS were halved every 8 wk. Airway hyperresponsiveness (AHR) to a bronchial provocation test (BPT) with histamine was measured at baseline. AHR to BPT with mannitol, spirometry, exhaled nitric oxide (eNO), and, in 31 subjects, sputum inflammatory cells were measured at baseline and at monthly intervals. Thirty-nine subjects suffered an asthma exacerbation. Seven subjects were successfully weaned off ICS. Using a Kaplan- Meier survival analysis, the significant predictors of a failure of ICS reduction were being hyperresponsive to both histamine and mannitol at baseline (p = 0.039), and being hyperresponsive to mannitol during the dose-reduction phase of the study (p = 0.02). Subjects older than 40 yr of age tended to be at greater risk of ICS reduction failure (p = 0.059). Response to mannitol and percentage sputum eosinophils were significantly greater before a failed ICS reduction than before the last successful ICS reduction, whereas there were no significant differences in symptoms, spirometry, or eNO. These findings suggest that documentation of patient's AHR or sputum eosinophils may be useful in guiding the reduction of ICS doses.