Red cells transfusion in the patient with acute hemorrhage, must be evaluated in a risk/benefit rate context. The present tendencies appoint that the use of the hematocrit "magic" number is unsafe and uncertain to decide a red cell transfusion. We have conducted a prospective randomized and controlled trial in 60 patients with acute digestive hemorrhage without haemodynamic failure. We realized two groups: 1) control group: the target of transfusion in these patients was the hematocrit value of > or = 28%. 2) treatment group: these patients were supported with normovolemic haemodilution with crystalloid solutions until a hematocrit value of 21%. All patients have endoscopic diagnosis and they went evaluated across the study with clinic and laboratory controls. Both groups were significative differences in the hematocrit value. We did not see differences between the groups in the hospital stay neither the rate of organs failure. We find difference between the groups in the amount of red cell units (0.61 +/- 0.87 vs. 2.14 +/- 1.10; treatment and control respectively, P < 0.001). The APACHE score was greater in the treatment group. This supports that the oldest patients, who probably have least physiologic reserve, could be treated without complications. Acute hemorrhage-normovolemic haemodilution-digestive hemorrhage transfusion.