Background Corticosteroids have been used to treat head injuries for more than 30 years. In 1997, findings of a systematic review suggested that these drugs reduce risk of death by 1-2%. The CRASH trial—a multicentre international collaboration—aimed to confirm or refute such an effect by recruiting 20 000 patients. In May, 2004, the data monitoring committee disclosed the unmasked results to the steering committee, which stopped recruitment. Methods 10 008 adults with head injury and a Glasgow coma score (GCS) of 14 or less within 8 h of injury were randomly allocated 48 h infusion of corticosteroids (methylprednisolone) or placebo. Primary outcomes were death within 2 weeks of injury and death or disability at 6 months. Prespecified subgroup analyses were based on injury severity (GCS) at randomisation and on time from injury to randomisation. Analysis was by intention to treat. Effects on outcomes within 2 weeks of randomisation are presented in this report. This study is registered as an International Standard Randomised Controlled Trial, number ISRCTN74459797. Findings Compared with placebo, the risk of death from all causes within 2 weeks was higher in the group allocated corticosteroids (1052 [21·1%] vs 893 [17·9%] deaths; relative risk 1·18 [95% CI 1·09-1·27]; p=0·0001). The relative increase in deaths due to corticosteroids did not differ by injury severity (p=0·22) or time since injury (p=0·05). Interpretation Our results show there is no reduction in mortality with methylprednisolone in the 2 weeks after head injury. The cause of the rise in risk of death within 2 weeks is unclear.