BACKGROUND: There is no agreement on how a bronchodilator response should be expressed. Ideally, the index used should be able to distinguish asthma from chronic obstructive lung disease and be independent of initial FEV1. METHODS: Two hundred and seventy four adults (aged 18-60 years) outpatients with obstructive airways disease were studied. Patients were divided into syndrome groups on the basis of a standardised history: asthma (n = 99), asthmatic bronchitis (n = 88), and chronic obstructive lung disease (n = 51); 36 subjects could not be attributed to any subgroup. FEV1 was measured before and 20 minutes after inhalation of 1000 micrograms terbutaline. Different expressions of bronchodilator response (delta FEV1) were compared with respect to their dependence on initial FEV1 and their efficacy in separating subjects with asthma from those with chronic obstructive lung disease. delta FEV1 was expressed as a percentage of initial FEV1 (delta FEV1%init), absolute value (delta FEV1), percentage of predicted FEV1 (delta FEV1%pred), standardised residual (delta SR-FEV1), and percentage of maximal possible increase (delta FEV1%[pred-init]). RESULTS: delta FEV1%init was more dependent on initial FEV1 (p = -0.405) than delta FEV1 (r = -0.145), delta FEV1%pred (r = -0.166), and delta SR-FEV1 (r = -0.127). delta FEV1%[pred-init] reached infinity when initial FEV1 approached predicted levels. delta FEV1%pred had a higher likelihood ratio (1.71) for separating patients with asthma from those with chronic obstructive lung disease than other expressions of bronchodilator response. Asthmatic patients had larger mean bronchodilator responses than patients in other subgroups; this difference was largest for delta SR-FEV1 (F = 9.19) and delta FEV1%pred (F = 9.03); it was much smaller for delta FEV1%init (F = 5.89). Despite significant differences in mean response, there was a large overlap of individual responses between diagnostic subgroups. The bronchodilator response was continuously and unimodally distributed for all expressions. CONCLUSIONS: delta FEV1%pred appears to be the most useful method of expressing bronchodilator response, both for clinical and for research purposes. Reversibility of airways obstruction in response to a bronchodilator is a continuous variable and not a dichotomous triat. Any cut off level of a "positive" bronchodilator response is therefore arbitrary.