Administration of BCNU into the carotid artery as treatment for malignant astrocytomas has produced retinal and brain toxicity. It is unclear whether the BCNU diluent, ethanol, is the cause of the toxicity, but the elimination of the ethanol is an attractive possibility. Clinically, decreasing the ethanol from 2.0 ml/100 mg BCNU to 0.75 ml/100 mg BCNU resulted in a marked decrease in eye toxicity. To simulate this clinical situation, three 500-mg solutions of BCNU, ethanol, and saline were prepared, decreasing the ethanol concentration from 3.0 ml to 2.0 ml to 0.75 ml/100 mg BCNU. The amount of BCNU recovered in vitro after simulated clinical administration of the three solutions decreased from 84.9% to 38.3% as the diluent decreased. Therefore, drug delivery at a fixed BCNU dose will decrease with the amount of ethanol diluent used. The clinical decrease in eye toxicity must be partly attributed to a decrease in the amount of BCNU delivered to the retina. Simulated administration of a solution of 500 mg BCNU/150 ml of 5% dextrose in water (D5W) gave 83.7% BCNU recovery. The D5W gives solubility comparable to that provided by 3.0 ml ethanol to each 100 mg BCNU, and its use eliminates ethanol as a potential retinal and brain toxin.