This paper reviews the incidence of, and risk factors for, drug-induced encephalopathy and mortality (from all causes) during treatment with melarsoprol of 1083 patients with Trypanosoma brucei gambiense sleeping sickness in Nioki hospital (Zaire) between 1983 and 1990. Sixty-four patients (5.9%) developed encephalopathy and 62 (5.7%) died: 43 from reactive encephalopathy and 19 from other causes. Univariate and multivariate analyses showed that the administration of prednisolone reduced significantly the incidence of encephalopathy and mortality during treatment, especially in patients with trypanosomes observed in the cerebrospinal fluid (CSF) and/or with a CSF white blood cell (WBC) count of 100 or more per mm3. The risk of encephalopathy was associated more strongly with the CSF WBC count than with the presence of CSF trypanosomes. In the subgroup of patients with a CSF WBC count of 100 or more mm3, changing the melarsoprol regimen to 3 series of 3 injections instead of 3 series of 4 injections halved the mortality rate during treatment. Treatment of patients who do develop reactive encephalopathy with the heavy metal chelator dimercaprol, in addition to intravenous steroids and anticonvulsants, may be harmful. The data suggest that a further reduction of the total dose of melarsoprol may decrease toxicity without jeopardizing efficacy.