Most children with asthma can be treated successfully with low-to-moderate doses of inhaled corticosteroid and long-acting beta-2 agonist. Those that fail to respond are a heterogeneous group. We propose that the nature and type of any steroid-resistant inflammation, the extent of any persistent airflow limitation and the extent of bronchial hyper-reactivity should be determined separately to allow a rational treatment approach to these children, rather than the haphazard advice of many current guidelines. Reasons for persistent difficult asthma include persistent eosinophilic inflammation, non-eosinophilic inflammation, airway reactivity without residual inflammation and persistent airflow limitation. We propose a protocol that uses non-invasive and invasive (bronchoscopic) methods to document the response to systemic steroids (depot triamcinolone). The aim of the protocol is to determine an individualised treatment plan; for example, cyclosporin for persistent eosinophilic inflammation, azithromycin for persistent neutrophilic inflammation and continuous subcutaneous terbutaline if there is airway reactivity without residual inflammation. Multi-centre studies are required to test the utility of this approach.