Childhood anaemia is a major public health problem in malaria holoendemic areas. We assessed the effects of antimalarial treatment in an area with drug-resistant falciparum malaria on haemoglobin levels in small children by applying the 1996 World Health Organization in vivo method for the evaluation of standard chloroquine treatment at the community level. In Fukayosi village, coastal Tanzania, 117 children aged 5-36 months with clinical malaria episodes were treated with chloroquine syrup (25 mg/kg). Early treatment failure (ETF) occurred in 20% and late treatment failure (LTF) in 22% of cases. Age > 1 year and malnutrition were protective factors against ETF. The evidence that chloroquine treatment could not prevent an exacerbation of anaemia was (i) the fact that the fall in haemoglobin level after 72 h was significantly greater in ETF than in children with LTF and an adequate clinical response, and (ii) the absence of any haematological improvement at follow-up in children receiving chloroquine alone, even in true treatment successes. In contrast, pyrimethamine/sulfadoxine administered to treatment failures improved the haemoglobin level significantly > 21 d after treatment started (mean difference 14 g/L, 95% confidence interval 2.1-27). We conclude that, when chloroquine treatment of childhood malaria is associated with a 20% ETF rate, the haemoglobin response is unsatisfactory and there is a need to change the recommended first-line treatment.