The efficacy of corticosteroids in chronic airflow limitation is variable and nonpredictable. We believe that the corticosteroid-induced improvement seen in a few subjects with chronic airflow limitation indicates coexistent bronchial asthma. Therefore, we hypothesize that by carefully excluding all features (other than corticosteroid response), either objective or subjective, that might be consistent with coexisting asthma, we could predict non-response to corticosteroids. Twenty patients with chronic airflow limitation related to cigarette smoking were chosen. Features that might suggest coexisting asthma were carefully excluded; these included rhinitis, nocturnal symptoms, atopy, large response to bronchodilator, and sputum or blood eosinophilia. Eighteen of these 20 subjects completed a double-blind random-order cross-over placebo-controlled trial of 30 mg of prednisone for 14 d (with a 14 d washout between treatments). There were no significant differences, individually or collectively, in objective or subjective determinations. Based on a greater than or equal to 250 mL improvement in FEV1, there were two placebo responders and two prednisone responders. Although it may be difficult to predict corticosteroid response in chronic airflow limitation, it appears that careful exclusion of 'asthmatic tendencies' may predict nonresponse to corticosteroids.