Dexamethasone was compared with placebo in a double-blind, crossover, randomised study of infants with severe bronchopulmonary dysplasia who had required mechanical ventilation for at least four weeks, despite treatment with diuretics, methylxanthines, bronchodilators, fluid restriction, nutritional supplementation, and ligation of the patent ductus arteriosus when indicated. Gestational age ranged from 27 to 33 weeks and birth weight from 800 to 1730 g. Patients received dexamethasone (0·5 mg/kg/day) or normal saline for the first 3 days, then treatment was crossed over for the next 3 days. The study was terminated when sequential analysis showed that all six patients had improved during dexamethasone therapy. Significant improvements were seen in ventilator-determined respiratory rate, peak inspiratory pressure, fractional inspired oxygen concentration, and alveolar arterial oxygen gradients (p<0·05). Although dexamethasone hastened weaning from mechanical ventilation, infection occurred in a substantial proportion of patients.