[en] Background It has been claimed that exhaled nitric oxide (FeNO) could be regarded as a surrogate marker for sputum eosinophil count in patients with asthma. However, the FeNO threshold value that identifies a sputum eosinophil count >/=3% in an unselected population of patients with asthma has been poorly studied. Methods This retrospective study was conducted in 295 patients with asthma aged 15-84 years recruited from the asthma clinic of University Hospital of Liege. Receiver-operating characteristic (ROC) curve and logistic regression analysis were used to assess the relationship between sputum eosinophil count and FeNO, taking into account covariates such as inhaled corticosteroids (ICS), smoking, atopy, age and sex. Results Derived from the ROC curve, FeNO >/=41 ppb gave 65% sensitivity and 79% specificity (AUC=0.777, p=0.0001) for identifying a sputum eosinophil count >/=3%. Using logistic regression analysis, a threshold of 42 ppb was found to discriminate between eosinophilic and non-eosinophilic asthma (p<0.0001). Patients receiving high doses of ICS (>/=1000 mug beclometasone) had a significantly lower FeNO threshold (27 ppb) than the rest of the group (48 ppb, p<0.05). Atopy also significantly altered the threshold (49 ppb for atopic vs 30 ppb for non-atopic patients, p<0.05) and there was a trend for a lower threshold in smokers (27 ppb) compared with non-smokers (46 ppb, p=0.066). Age and sex did not affect the relationship between FeNO and sputum eosinophilia. When combining all variables into the logistic model, FeNO (p<0.0001), high-dose ICS (p<0.05) and smoking (p<0.05) were independent predictors of sputum eosinophilia, while there was a trend for atopy (p=0.086). Conclusion FeNO is able to identify a sputum eosinophil count >/=3% with reasonable accuracy and thresholds which vary according to dose of ICS, smoking and atopy.