To evaluate the efficacy of dexamethasone for treatment of primary supratentorial intracerebral hemorrhage, we studied 93 patients 40 to 80 years old, using a double-blind randomized block design. After the subjects were stratified according to their level of consciousness (Glasgow Coma Scale), those with objectively documented primary supratentorial intracerebral hemorrhage were randomly assigned to either dexamethasone or placebo. For ethical reasons, three interim analyses were planned, to permit early termination of the trial if one study group did better than the other. During the third interim analysis, the death rate at the 21st day was identical in the two groups (dexamethasone vs. placebo, 21 of 46 vs. 21 of 47; chi-square = 0.01, P = 0.93). In contrast, the rate of complications (mostly infections and complications of diabetes) was much higher in the dexamethasone group (chi-square = 10.89, P less than 0.001), leading to early termination of the study. In the light of the absence of a demonstrable beneficial effect and the presence of a significant harmful effect, current practices of using dexamethasone for treatment of primary supratentorial hemorrhage should be reconsidered.