A prospective randomised study was performed on 25 children aged 1.4 to 15.8 years with severe head injury (Glasgow Coma Scale less than or equal to 7) to determine the clinical effectiveness and the impact on endogenous cortisol production of high-dose steroid therapy. Thirteen patients (group 1) received dexamethasone 1 mg/kg/day during the first 3 days and 12 (group 2) not. All patients were treated with a standardized regimen. Urinary free cortisol was measured by radioimmunoassay, and the clinical data were recorded at hourly intervals. Outcome was assessed 6 months later using the Glasgow Outcome Scale. We found a higher frequency of bacterial pneumonias in the dexamethasone-treated patients (7/13 versus 2/12). Group 1 showed a suppression of endogenous cortisol production from day 1 to day 6. In group 2, mean free cortisol was up to 5-fold higher than under basal conditions. The results in group 2 showed that the endogenous steroid production reacts adequately to the stress of severe head injury. It probably is sufficient to elicit maximum glucocorticoid effects. There was no other statistically significant difference in the clinical and laboratory data between the two groups. We conclude that dexamethasone in high doses suppresses endogenous cortisol production up to 6 days and may increase the risk of bacterial infection without affecting the outcome or the clinical and laboratory data.