Background: Although inhaled corticosteroids (ICSs) are effective in preventing deterioration in asthma control, at least half of subjects with mild-to-moderate asthma will remain stable when these agents are discontinued.

Objective: We sought to determine whether noninvasive markers of inflammation predict which individuals maintain asthma control after discontinuation of ICSs.

Methods: We analyzed data obtained from 164 subjects with mild-to-moderate asthma who participated in a 16-week trial comparing the effects of continued ICS use with the effects of a switch to salmeterol or placebo.

Results: In comparison with continued ICS use, a switch to salmeterol or placebo was associated with increased rates of asthma deterioration over 16 weeks (9.3% vs 24.1% and 37.5%, respectively; P = .04 and P < .001, respectively). We found that neither exhaled nitric oxide nor methacholine PC 20 , when measured at randomization or 2 weeks after randomization, were significant predictors of subsequent asthma control in subjects who discontinued ICSs. However, both induced sputum eosinophil counts measured 2 weeks after a switch from ICS to placebo and changes in sputum eosinophil counts from before cessation of ICSs to after a switch to placebo predicted subsequent asthma deterioration (area under the receiver-operating characteristic curve, 0.771 [ P < .001] and 0.825 [ P < .001], respectively).

Conclusion: On the basis of a model treatment strategy, we estimate that allocating subjects to ICS therapy on the basis of changes in sputum eosinophil counts after a trial discontinuation could allow 48% of subjects with mild-to-moderate asthma to discontinue ICS therapy without an increased risk of asthma deterioration over a period of at least 14 weeks.