Methods to examine sputum for indices of airway inflammation are evolving. We have examined the repeatability and the validity of an improved method to measure sputum cells and fluid-phase eosinophil cationic protein (ECP), major basic protein (MBP), eosinophil-derived neurotoxin (EDN), albumin, fibrinogen, tryptase, and interleukin-5 (IL-5). Sputum was induced with hypertonic saline twice within 6 d in 10 healthy subjects, 19 stable asthmatics, and 10 smokers with nonobstructive bronchitis. The method included the processing of freshly expectorated sputum separated from saliva, treatment with a fixed proportion of dithiothreitol 0.1% followed by Dulbecco's phosphate-buffered saline, making cytospins, and collecting the supernatant. The reproducibility of measurements, calculated by the intraclass correlation coefficient, was high for all indices measured with the exception of total cell counts and proportion of lymphocytes. Asthmatics, in comparison with healthy subjects and smokers with bronchitis, had a higher proportion of sputum eosinophils (median percent 5.2 versus 0.5 and 0.3), metachromatic cells (0.3 versus 0.07 and 0.08), ECP (1,040 micrograms/L versus 288 and 352), MBP (1,176 micrograms/L versus 304 and 160), and EDN (1,512 micrograms/L versus 448 and 272). Asthmatics differed from healthy subjects, but not from smokers with bronchitis, in the proportion of neutrophils (46.9% versus 24.1%), albumin (704 versus 288 micrograms/mL), and fibrinogen (2,080 versus 440 ng/mL). Smokers with bronchitis showed a trend for a higher neutrophil count and levels of albumin and fibrinogen than healthy subjects. The proportion of sputum eosinophils correlated positively with ECP, MBP, EDN, albumin and fibrinogen levels, and metachromatic cell counts correlated with tryptase. In asthmatics, IL-5 correlated with eosinophil counts. There was a significant negative correlation between sputum indices and expiratory flows and methacholine PC20. Thus, the methods of measuring cell and fluid phase markers in induced sputum used in this study are reproducible and valid. They can therefore be used to reliably measure these indices of airway inflammation.