Measurement of nitric oxide in exhaled air is a noninvasive method to assess airway inflammation in asthma. This study was undertaken to establish the reference range of exhaled NO in healthy school-aged children and to determine the influence of ambient NO, noseclip and breath-holding on exhaled NO, using an off-line balloon sampling method. All children attending a primary school (age range 8-13 yrs) underwent NO measurements on two occasions with high and low ambient NO. Each time, the children performed four expiratory manoeuvres into NO-impermeable balloons, with and without 10 s of breath-holding and with and without wearing a noseclip. Exhalation flow and pressure were not controlled. NO was measured within 4 h after collection, by means of chemiluminescence. All children completed a questionnaire on respiratory and allergic disorders, and performed flow/volume spirometry. With low ambient NO, the mean exhaled NO value of 72 healthy children with negative questionnaires and normal lung function was 5.1 +/- 0.2 parts per billion (ppb) versus a mean of 6.8 +/- 0.3 ppb in the remaining 49 children with positive questionnaires for asthma and allergy, and/or recent symptoms of cold (p=0.001). Exhaled and ambient NO were significantly related, especially with ambient NO > 10 ppb (r = 0.86, p=0.0001 versus r=0.34, p=0.004 for ambient values <10 ppb). The use of a noseclip, with low ambient NO and without breath-holding, caused a small decrease in exhaled NO values (p=0.001). The effect of breath-holding on exhaled NO depended on ambient NO. With ambient NO > 10 ppb, exhaled NO decreased, whereas with ambient NO < 10 ppb, exhaled NO increased after 10 s breath-hold. It is concluded that off-line sampling in balloons is a simple and, hence, attractive method for exhaled nitric oxide measurements in children which differentiates between groups with and without self-reported asthma, allergy and colds, when ambient nitric oxide is < 10 parts per billion. Wearing a noseclip and breath-holding affected measured values and should, therefore be standardized or, preferably, avoided.