OBJECTIVE: In 2002-03, the Republic of the Congo increased the threshold separating stage 1 and 2 cases of human African trypanosomiasis (HAT) from a cerebrospinal fluid (CSF) white cell count of 5 cells/mm(3) to 10 cells/mm(3). We aimed to assess whether the increased threshold of 10 cells/mm(3) is a safe indicator of stage 2 disease. METHODS: We assessed patients treated for stage 1 HAT caused by Trypanosoma brucei gambiense in the Republic of the Congo between April 2001 and April 2005. Patients with 0-10 cells/mm(3) in CSF were classed as stage 1 and treated with pentamidine. Patients with CSF of > 10 cells/mm(3) were classed as stage 2 and treated with either melarsoprol or eflornithine. We did a retrospective analysis of all patients treated after the September 2002 increase in threshold for classification of HAT disease stage 2, and who were eligible for at least 1 year of follow-up. Primary outcome was survival without death or relapse within 1 year of discharge. Risk factors for treatment failure, in particular CSF white cell count on diagnosis, were assessed. FINDINGS: Between September 2002 to April 2004, 692 patients eligible for our analysis were treated with pentamidine. All were discharged alive. Relapse rate was 5% (n = 33). The only identified risk factor for relapse was a CSF white cell count of 6-10 cells/mm(3) rather than 0-5 cells/mm(3) (adjusted hazard ratio 3.27 (95% confidence interval, 1.52-7.01); P = 0.002). CONCLUSION: A threshold of 5 white cells/mm(3) in CSF is safer than 10 cells/mm(3) to determine stage 2 HAT and reduce risk of relapse.