OBJECTIVE: To compare the effectiveness of melarsoprol and eflornithine in treating late-stage Gambian trypanosomiasis in the Republic of the Congo. METHODS: We analysed the outcomes of death during treatment and relapse within 1 year of discharge for 288 patients treated with eflornithine, 311 patients treated with the standard melarsoprol regimen and 62 patients treated with a short-course (10-day) melarsoprol regimen between April 2001 and April 2005. FINDINGS: A total of 1.7% (5/288) of patients treated with eflornithine died compared with 4.8% (15/311) of those treated with standard melarsoprol and 6.5% (4/62) of those treated with short-course melarsoprol. Patients treated with eflornithine tended to be younger and were more likely to have trypanosomes or higher white blood cell counts in their cerebrospinal fluid. The cumulated incidence of relapse among patients who attended at least one follow-up visit 1 year after discharge was 8.1% (11/136) for those treated with eflornithine, 14% (36/258) for those treated with standard melarsoprol and 15.5% (9/58) for those treated with shortcourse melarsoprol. In a multivariate analysis, when compared with eflornithine, standard melarsoprol was found to be a risk factor for both death (odds ratio (OR) = 2.87; 95% confidence interval (CI) = 1.03-8.00) and relapse (hazard ratio (HR) = 2.47; 95% CI = 1.22-5.03); when compared with eflornithine, short-course melarsoprol was also found to be a risk factor for death (OR = 3.90; 95% CI = 1.02-14.98) and relapse (HR = 6.65; 95% CI = 2.61-16.94). CONCLUSION: The effectiveness of melarsoprol treatment appears to have diminished. Eflornithine seems to be a better first-line therapy for treating late-stage Gambian trypanosomiasis in the Republic of the Congo.