The safety of sputum induction and the reproducibility of measurements in induced sputum in multicenter studies is unknown. We examined the safety of sputum induction in a two-visit, six-center study in 79 subjects with moderate to severe asthma (mean ± SD FEV₁ 71 ± 12% predicted, 67% taking inhaled corticosteroids). In addition, we compared the reproducibility of markers of inflammation in induced sputum with the reproducibility of the FEV₁ and the methacholine PC₂₀. The FEV₁ decreased ⩾ 20% from the postbronchodilator baseline in 14% of all subjects and in 25% of subjects whose initial prebronchodilator baseline was 40 to 60% of predicted. All subjects responded promptly to additional albuterol treatment, and no subject developed refractory bronchoconstriction requiring treatment other than reversal of bronchospasm in the study laboratory. The reproducibility of measurements of the eosinophil percentage, eosinophil cationic protein, tryptase, and methacholine PC₂₀ were similar (concordance correlation coefficients of 0.74, 0.81, 0.79, and 0.74, respectively), without any significant among-center effect. We conclude that sputum induction can be performed safely in subjects with moderate to severe asthma in multicenter clinical trials when carried out under carefully monitored conditions. Importantly, we demonstrate that measurement of markers of inflammation in induced sputum is as reproducible as methacholine PC₂₀ and should prove useful in the assessment of airway inflammation in multicenter clinical trials.