Sputum induction is a noninvasive method to evaluate airway inflammation. We investigated whether it can be safely and successfully performed in patients with severe, difficult-to-control asthma, and whether the patients at risk can be identified. Ninety-three severe asthmatics were included, all symptomatic despite inhaled corticosteroids (> or = 1,600 microg/d) and long-acting beta(2)-agonists > 1 yr. Patients with a postbronchodilator FEV(1) < 1 L and < 50% predicted were excluded from participation. Sputum induction was performed according to a strict protocol, using 0.9%, 3.0%, and 4.5% NaCl inhalation. In 74% (CI: 64 to 83%) of patients an adequate sputum sample could be obtained. Twenty-two percent (CI: 14 to 33%) developed excessive bronchoconstriction (decrease in FEV(1) > 15% from baseline) despite the continuing use of long-acting bronchodilators and pretreatment with 400 microg salbutamol. The decrease in FEV(1) was associated with increased use of rescue short-acting beta(2)-agonists in the previous 2 d (r(s) = 0.51, p = 0.002), lower postbronchodilator FEV(1) (r(s) = -0.31, p = 0.004), and lower provocative concentration of histamine causing a 20% reduction in FEV(1) (PC(20)) (r(s) = -0.52, p < 0.001). Recent use of short-acting beta(2)-agonist increased the risk for excessive bronchoconstriction 10.2-fold (CI: 1.2 to 109.8). In conclusion, sputum induction can be safely and successfully performed in patients with severe, difficult-to-control asthma if a standardized protocol is used. However, severe bronchoconstriction may occur despite regular use of long-acting beta(2)-agonist and pretreatment with salbutamol 400 microg. In particular, patients who have used additional short-acting beta(2)-agonists as rescue medication during the days preceding the induction, are at high risk.