OBJECTIVE: Evaluate the use of normovolaemic haemodilution in cervico-facial oncologic surgery. STUDY DESIGN: Prospective, randomised, simple blinded study. PATIENTS AND METHODS: 38 ASA I and II patients were studied: a control group (n = 21) and a haemodilution group (n = 17) in whom 5.5-8 mL.kg-1 blood were withdrawn before induction, replaced by an equivalent amount of colloids. In both groups, the transfusional strategy was to keep the haemoglobin level above 100 g.L-1 throughout the procedure and the recovery phase, using in priority the autologous blood in the haemodilution group. Blood losses during surgery were evaluated by weighing the sponges and by measuring the aspirated blood. RESULTS: Demographic and anaesthetic data, and blood losses were similar. The haemoglobin levels dropped significantly in the haemodilution group (138 +/- 10 g.L-1 to 107 +/- 11 g.L-1) as well as in the control group (131 +/- 11 g.L-1 to 110 +/- 10 g.L-1). Infectious complications were slightly higher in the haemodilution group, although this difference did not reach the level of significance. CONCLUSION: Normovalaemic haemodilution does not seem to be indicated in cervico-facial and ENT oncologic surgery.