In a placebo-controlled trial of co-trimoxazole prophylaxis in Côte d'Ivoire, neutropenia was the most frequent short-term side effect. The long-term incidence of neutropenia in sub-Saharan African adults receiving co-trimoxazole has never been reported. We followed a prospective cohort of HIV-infected adults receiving co-trimoxazole (sulphamethoxazole 800 mg/trimethoprim 160 mg daily) in Abidjan. Grades of neutropenia were successively defined as at least one absolute neutrophil count (ANC) of: <1500/mm(3) (severity grade >/=1), <1000/mm(3) (grade >/=2), <750/mm(3) (grade >/=3) or <500/mm(3) (grade 4). In total, 533 adults were followed-up during 1450 person-years, with a total of 3154 ANCs. The probability of remaining free of neutropenia at 48 months was 0.29 (95% CI 0.23-0.34) for grade >/=1, 0.64 (95% CI 0.60-0.71) for grade >/=2, 0.82 (95% CI 0.77-0.86) for grade >/=3 and 0.96 (95% CI 0.93-0.99) for grade 4. The only factor significantly associated with a higher rate of all grades of neutropenia was a low baseline CD4 count. There was no association between any grade of neutropenia and the global risk of serious morbidity during the study period. In adults receiving co-trimoxazole in Abidjan, mild neutropenia is a common observation with no evidence of negative clinical consequences. The consequences of associating co-trimoxazole with other haematotoxic drugs should be carefully assessed.