We performed a prospective, randomized trial of two different strategies for postoperative packed red blood cell replacement in 39 autologous blood donors undergoing elective myocardial revascularization. The "liberal" group received blood to achieve a hematocrit value of 32%, and the "conservative" group received transfusions for a hematocrit value less than 25%. Although the groups had significantly different mean hematocrit values from the fourth postoperative hour (28.7% versus 31.2%) through the fifth postoperative day (28.4% versus 31.3%), there were no significant differences in fluid requirement, hemodynamic parameters, or hospital complications. Significantly fewer units of packed cells were required in the conservatively transfused group (20 units/20 patients) compared with the liberally transfused group (37 units/18 patients) (p = 0.012). Exercise tests were performed on the fifth and sixth postoperative days, with a transfusion being given to the conservative group between tests. Although a significant improvement in exercise endurance occurred in the conservative group receiving a transfusion (p = 0.008), no significant difference in duration or degree of exercise was demonstrated between the two groups on either day. In comparing these two groups of profoundly anemic patients, we identified no adverse consequence associated with the greater degree of hemodilution and could identify no correlation between hematocrit value and exercise capacity. We conclude that although the limits of hemodilution are still poorly defined, postoperative blood transfusion in revascularized patients should be guided by clinical indications and not by specific hematocrit values.