Background: Knowing the safety profile of anti-malarial treatments in routine use is essential; millions of patients receive now artemisinin combination therapy (ACT) annually, but the return on information through current systems is as yet inadequate. Cohort event monitoring (CEM) is a WHO (World Health Organization)-recommended practice; testing its performance and feasibility in routine practice in malaria-endemic is important. Methods. A nine-year CEM-based study of the safety of artesunate-amodiaquine (ASAQ) at five peripheral health facilities in a rural district of South-western Senegal. Staff (nurses, health workers) were trained to collect actively and systematically information on the patient, treatment and events on a purposely designed questionnaire. The occurrence and severity of events was collected before, during and after treatment up to 28 days in order to generate information on all adverse events (AEs) as well as treatment-emerging signs/symptoms (TESS). Laboratory tests (haematology, liver and renal) was planned for at least 10% of cases. Results: During 2001-2009, 3,708 parasitologically-confirmed malaria cases (mean age = 16.0 ± 12.7 years) were enrolled (26% and 52% of all and parasitologically-confirmed ASAQ treatments, respectively). Treatment was supervised in 96% of cases. Products changed over time: 49% were a loose combination of individually-packaged products (available 2001-03), 42% co-blistered products (2004-09) and 9% a fixed-dose co-formulation (2006-09); dosing was age-based for 42%, weight-based for 58%. AS and AQ were correctly dosed in 97% and 82% of cases with the loose and 93% and 86% with the fixed combination, but only 50% and 42% with the co-blistered product.Thirty-three per cent (33%) of patients had at least one sign/symptom pre-treatment, 12% had at least one AE and 9% a TESS (total events 3,914, 1,144 and 693, respectively). AEs overestimated TESS by 1.2-2 fold (average 1.7). Changes in laboratory value were insignificant. Over-dosing more than doubled the risk of TESS, though statistical significance was reached only during 2003-2007. The incidence of serious events (including death) was five per thousand. Conclusions: The study was successful in quantifying and characterizing known reactions and has benchmarking value. Health staff performance varied. Investments in training, motivating and providing a quality control system would be needed. The study proved that a CEM-based system is feasible in this setting but more research is needed to assess whether it is sustainable and what conditions would make it cost-effective, including the amount and quality of data generated, and the use thereof for decision-making.