Efficacy of a Simple Intraoperative Transfusion Algorithm for Nonerythrocyte Component Utilization after Cardiopulmonary Bypass

After obtaining approval from the Mayo Medical Center Institutional Review Board (Rochester, MN) and written informed consent, adult men and nonpregnant adult women scheduled for elective cardiac surgery requiring CPB were randomized using a computer-generated randomization list with a block size of four to one of two groups. The two groups were a control group (transfusion support guided by empiric clinical judgment) and a protocol group (transfusion guided by an algorithm based on coagulation test results. Intraoperatively after CPB, the anesthesiologist and surgeon assessed extent of bleeding; abnormal microvascular bleeding (diffuse oozing with no visible clot) was determined by inspection of the operative field. Once it was determined that the patient had abnormal bleeding, transfusion therapy was determined by group designation. The patients were randomized on enrollment in the study. The surgeons and anesthesiologists were not made aware of which group the patients were placed in until after they decided that the patient had abnormal bleeding after CPB and they felt the patient needed to have transfusion of non-erythrocyte components. Therefore, the people making the transfusion decisions were blinded to group designation of the patients until after the determination of abnormal bleeding after CPB.

All patients received a moderate-dose opioid-based anesthetic technique, supplemented with benzodiazepines, muscle relaxants, and inhalational anesthetic agents. CPB was conducted with a Univox membrane oxygenator (Bentley Inc., Irvine, CA) and a Sarns 9000 CPB machine using a roller head (Sarns Inc., Ann Arbor, MI) at a flow of 2.4 l · min⁻¹· m⁻². The CPB circuit was primed with 1.5 l of balanced salt solution, 10 mEq sodium bicarbonate (NaHCO³), and 12.5 g mannitol. Porcine heparin was administered to patients as follows: an initial dose was given consisting of a bolus of 300 units/kg and an oxygenator priming dose of 10,000 units. Additional heparin (5,000 units) was administered when the celite activated clotting time (ACT) was less than 450 s or 750 s for patients on aprotinin therapy. After discontinuation of CPB, the initial protamine sulfate dose was 0.013 mg per unit of heparin administered. Heparin neutralization was regarded as adequate if the postprotamine ACT value was within 10% of the preheparin ACT value. Additional protamine (20–50 mg) was added at the discretion of the attending anesthesiologist if the ACT had not returned to that range, with the goal of returning the ACT to within 10% of the preheparin value. The designation of whether the patient had abnormal bleeding or not occurred 10 min after return of the ACT to within 10% of the preheparin value. Intraoperative blood salvage and reinfusion of shed mediastinal blood was used in all cases. Prophylactic antifibrinolytic therapy, aprotinin, tranexamic acid, and epsilon amino-caproic acid were used at the discretion of the attending anesthesiologist. Patients were not excluded if they received preoperative aspirin or antiplatelet therapy.

The control group received transfusion of blood products based on the clinician’s judgment with or without guidance from laboratory tests. Twelve anesthesiologists and six surgical teams participated in the study. In the control group, the decisions for transfusion of blood components (varying combinations of platelets, fresh frozen plasma, or cryoprecipitate) were made by the attending anesthesiologist but could be empirical and precede the availability of any coagulation tests.

Participants in the treatment algorithm group had their transfusion therapy guided by an algorithm based on coagulation tests (fig. 1). The laboratory values used to guide therapy in the transfusion algorithm were determined from previous studies of patients undergoing cardiac surgery requiring CPB by performing receiver operator characteristic curve analysis at the following time point: in the operating room 10 min after protamine administration and return of the ACT to within 10% of baseline values. 23,24The receiver operator characteristic curve analysis allowed us to determine which tests and test values produced greatest sensitivity, specificity, and accuracy for differentiating patients with abnormal bleeding from those without. Test values with the greatest accuracy were placed in our transfusion algorithm. Once it was determined that the patient was bleeding abnormally and the ACT had returned to within 10% of baseline, the laboratory tests for the algorithm were drawn. The algorithm directed transfusion of blood products, if any. Allogeneic platelets or desmopressin acetate were administered if the platelet count was less than 102,000/μl or the maximum amplitude of the thromboelastograph was less than 48 mm.

Allogeneic fresh frozen plasma was transfused if the prothrombin time was greater than 16.6 s or the activated partial thromboplastin time (APTT) was greater than 57 s. Cryoprecipitate was transfused if the fibrinogen was less than 144 mg/dl. The fibrinogen concentration test was a slow laboratory test and was used to guide the transfusion of cryoprecipitate in the second round of allogeneic blood transfusion, if necessary. If the patient continued to have abnormal bleeding by observation of the surgical field after the administration of allogeneic blood products, another round of laboratory tests for the algorithm were drawn. The algorithm directed transfusion of blood products, if any.

A subset of the control group, 25 patients, had the algorithm coagulation tests performed 10 min after protamine administration and normalization of the ACT. This was performed to compare coagulation test results between the two groups. The results of the coagulation tests in this subset were not provided to the clinician determining the type of and need for allogeneic blood transfusions.

The point-of-care tests used in the transfusion algorithm were the whole blood prothrombin time and APTT measured with the Biotrack 512 Coagulation Monitor (currently known as the CoaguChek Plus, Roche Diagnostics, Indianapolis, IN), a portable battery-powered unit that uses disposable plastic reagent cartridges. 23The results are obtained within 3–6 min and have an excellent correlation with the hospital laboratory prothrombin time and APTT. 25The platelet counts were determined by the Coulter MD II (Coulter Corp., Hialeah, FL), and results were available within 5–10 min. The whole blood thrombelastogram (Haemoscope Corporation, Morton Grove, IL) maximum amplitude 21results were available within 30 min. The fibrinogen concentration 26results were available within 1 h.

All patients had laboratory-based tests (prothrombin time, APTT, platelet count) performed on arrival in the ICU. Mediastinal drainage in the ICU was recorded for the first 4, 12, and 24 h postoperatively. All allogeneic blood use was recorded intraoperatively, in the first 24 h postoperatively, and for the entire hospitalization period. Whether the patients with abnormal bleeding required reoperation for exploration of their mediastinum for bleeding was documented along with the findings of the reoperation. Use of antifibrinolytic therapy and preoperative aspirin or antiplatelet therapy was recorded.

Power analysis was based on data from a previous retrospective study of patients who had undergone cardiac surgery at our clinic and a previous study by Despotis et al.  19This power analysis determined that we could detect (α= 0.05, 80% power) a change on the geometric scale from 6.4 to 1.6 units of allogeneic blood products transfused with 92 total patients having abnormal bleeding. We assumed 10% of our patients would develop microvascular bleeding, and thus planned to enroll up to 920 patients into the study.

We compared blood loss, autologous blood returned, and the number of allogeneic blood products (erythrocytes, fresh frozen plasma, platelets, and cryoprecipitate) administered between groups. The Fisher exact test was used to assess differences between groups for categoric variables. Differences between groups for continuous variables were assessed using the Wilcoxon rank sum test. Multiple regression analyses were used to assess differences in fresh frozen plasma and platelet transfusion between groups after adjusting for preoperative variables that were significantly different between groups. In these regression analyses, the dependent variables were number of units of fresh frozen plasma and platelets transfused in the operating room. Finally, multiple regression analyses were performed to assess the effect of wound drainage on amount of fresh frozen plasma and platelets transfused. In these analyses, the dependent variables were the 4-, 12-, and 24-h chest tube drainage, whereas the independent variables were the amount of fresh frozen plasma and platelets transfused in the operating room. P  values ≤ 0.05 were considered significant.

Among the 836 enrolled patients having elective cardiac surgery requiring CPB, 92 developed abnormal bleeding (11% incidence;fig. 2). Four of the patients initially randomized to the algorithm group were converted to the control group because of unavailability of study personnel. Thus, 41 patients were entered into the algorithm group, and 51 were entered into the control group.

The two groups were similar with regard to preoperative characteristics, except that 34.2% of patients in the algorithm group and 15.7% in the control group were receiving preoperative warfarin therapy (P = 0.05;table 1). There was no difference in intraoperative blood loss after CPB between the groups. No patients received epsilon amino-caproic acid or desmopressin acetate. There were no differences in coagulation test values, except for small differences in the preoperative platelet count (table 2). No patients had autologous blood available for surgery.

The percentage of patients who received different types of allogeneic blood transfusions in the operating room and the first 24 h in the ICU are shown in figure 3. The algorithm group received a median of 0 units (range, 0–7 units) of allogeneic fresh frozen plasma in the operating room after CPB compared with a median of 3 units (range, 0–10 units) in the control group (P = 0.0002). The median number of platelet units transfused in the operating room after CPB was 4 (range, 0–12) in the algorithm group compared with 6 (range, 0–18) in the control group (P = 0.0001). Our transfusion algorithm did not guide erythrocyte transfusions in the operating room, and there was no difference in the amount of erythrocyte units transfused in the operating room. The algorithm group received a median of 1 unit (range, 0–12 units) of allogeneic erythrocytes in the operating room after CPB compared with a median of 1 unit (range, 0–6 units) in the control group (P = 0.85).

During the first 24 h in the ICU, there were no differences in allogeneic platelet and fresh frozen plasma transfusions between the groups. The algorithm group received significantly less allogeneic erythrocytes, with a median of 0 units (range, 0–6 units) in the ICU compared with a median of 1 unit (range, 0–6 units) in the control group (P = 0.0028). There was no transfusion algorithm used in the ICU.

The amount of blood transfused to the algorithm group throughout the hospitalization was less than the control group. The algorithm group received a median of 2 units (range, 0–10 units) of allogeneic fresh frozen plasma throughout the hospitalization period compared with a median of 4 units (range, 0–75 units) in the control group (P = 0.0005). The median number of platelet units transfused throughout the hospitalization period was 6 (range, 0–18) in the algorithm group compared with 6 (range, 0–144) in the control group (P < 0.0001). The algorithm group received a median of 2 units (range, 0–9 units) of allogeneic erythrocytes throughout the hospitalization period compared with a median of 3 units (range, 0–70 units) in the control group (P = 0.039).

There was greater blood loss in the ICU for the control group compared with the algorithm group (table 3). The control group also had a greater incidence of surgical reoperation of the mediastinum for bleeding than the transfusion algorithm group (six control patients, zero algorithm patients;P = 0.032).

Since four patients in the algorithm group were moved to the control group because of the unavailability of study personnel, we repeated the analysis as an intention-to-treat analysis. Despite these patients not having their transfusion therapy guided by the algorithm, there still was a significant reduction in platelet and fresh frozen plasma transfusions in the operating room for the algorithm group relative to the control group. The algorithm group received a median of 2 units (range, 0–8 units) of allogeneic fresh frozen plasma in the operating room after CPB compared with a median of 3 units (range, 0–10 units) in the control group (P = 0.0019). The median number of platelet units transfused in the operating room after CPB was 6 (range, 0–12) in the algorithm group compared with 6 (range, 0–18) in the control group (P < 0.0001). The algorithm group received a median of 0 units (range, 0–6 units) of allogeneic erythrocytes in the first 24 h in the ICU compared with a median of 1 unit (range, 0–6 units) in the control group (P = 0.0056).

Because there were some differences between the groups in the preoperative warfarin use, preoperative platelet count, and the presence of other cardiac procedures, a multiple regression analysis was performed to adjust for these factors on transfusion outcome. The dependent variables were the amount of fresh frozen plasma and platelet units transfused in the operating room. The independent variables were preoperative warfarin use (0 = no, 1 = yes), preoperative platelet count, maze procedure (0 = no, 1 = yes), patent foramen ovale closure (0 = no, 1 = yes), aortic arch procedure (0 = no, 1 = yes), other cardiac surgical procedure (0 = no, 1 = yes), and group (0 = control, 1 = algorithm). After adjusting for preoperative warfarin use, preoperative platelet count, and other cardiac procedures performed, the algorithm group had significantly fewer units of fresh frozen plasma (P = 0.0001) and platelets (P = 0.0001) transfused in the operating room.

The amount of chest tube drainage was also significantly reduced in the algorithm group at all time points in the first 24 h in the ICU compared with the control group. Because chest tube drainage in the ICU might be affected by the amount of blood given in the operating room, a multiple regression analysis was performed to determine if the group differences remained significant after adjusting for the amount of fresh frozen plasma and platelet units transfused in the operating room. In this regression, the dependent variables were 4-, 12-, and 24-h chest tube drainage. The independent variables were fresh frozen plasma and platelet units transfused in the operating room. After adjusting for the fresh frozen plasma and platelet units transfused, there were no longer significant differences between the groups in 4-h chest tube drainage (P = 0.15), the 12-h chest tube drainage (P = 0.1), or the 24-h chest tube drainage (P = 0.14).

In the year 2000, approximately 500,000 patients in the United States underwent cardiac surgery using CPB. 27The majority of these patients (50–75%) likely received allogeneic blood transfusion. 17,22Between 10 and 20% of the approximately 12 million units of erythrocytes transfused in the United States annually are given to cardiac surgical patients at a direct societal cost of 0.2–0.4 billion dollars. 28,29Approximately 50% of the more than 7 million units of platelets transfused annually are to patients undergoing cardiac surgery, and use of platelet transfusion continues to grow.

Clinicians are limited in their ability to predict abnormal bleeding after CPB and identify patients who will require blood transfusion. In fact, the decision to transfuse cardiac surgery patients today in many institutions remains a highly subjective and relatively antiquated activity. There is large variability between institutions in transfusion behavior for cardiac surgery patients. 17Approximately 20% of blood transfusion in cardiac surgery patients are inappropriate. 9,17Specifically, 47% of platelets, 32% of fresh frozen plasma, and 15% of erythrocytes are considered inappropriate transfusions. 9The decision to transfuse blood should be made with relevant laboratory data, 8,10,11yet this may not be commonly performed. The most recent consensus conference by the American Society of Anesthesiologist stated, “the lack of data from prospective, randomized studies with adequate sample size, control groups, clinical outcome measurements, and other features of well designed clinical effectiveness research impedes development of evidence-based clinical practice guidelines for blood component therapy.”11We previously determined in our patient population the coagulation test values that would most effectively guide transfusion of blood products in our cardiac surgical patients, and these values were incorporated into a transfusion algorithm used in this study. 23,24Our prospective randomized study has demonstrated a significant reduction in allogeneic platelet and fresh frozen blood transfusions in the operating room in the algorithm group compared with the control group. Furthermore, there was a reduction in use of intraoperative combination therapy, transfusion of platelets with fresh frozen plasma or cryoprecipitate, from 90% in the control group down to 32% in the algorithm group (fig. 3). Use of the transfusion algorithm resulted in greater use of specific types of blood components solely or not at all. This represented a reduction in the use of unnecessary or inappropriate blood component transfusions. This reduction occurred in the area of greatest non-erythrocyte component use, the operating room (fig. 3).

An unanticipated but important finding of our study is that the lower number of coagulation product transfusions in the operating room in the algorithm group may have resulted in less bleeding in the ICU than the control group. This striking reduction in blood loss in the ICU after the use of the algorithm in the operating room compared with the control group was also found in another prospective randomized study. In this study, Despotis et al.  19reported that a coagulation test–based intraoperative transfusion algorithm in cardiac surgery patients who had microvascular (nonsurgical) bleeding was effective in reducing allogeneic blood product transfusion. In both our study and that by Despotis et al. , 19the more directed transfusion therapy may have corrected the hemostatic problem more effectively, which resulted in reduced bleeding in the ICU. In addition, the algorithm group had a significantly lower incidence of mediastinal reexploration for hemorrhage than the control group. Mediastinal reexploration for hemorrhage increases patients’ morbidity and mortality, and increases hospital costs. 12,13 

In the current study, we did not extend the transfusion algorithm to the ICU, and there were no differences in transfusion coagulation components in the ICU. There was, however, a reduction in erythrocyte transfusions in the algorithm group.

Outside of the operating room and the ICU, thromboelastography is not often used. We used the maximum amplitude as a point-of-care test of platelet function. Others have used this test to guide transfusion therapy. 20,21The maximum amplitude is influenced by platelet function, 30platelet count, 31and fibrinogen concentration. 32 

Four patients were reassigned from the algorithm group to the control group because of the unavailability of study personnel, and this may have introduced bias in our study. However, reanalysis of our data based on intention to treat resulted in similar significant reductions in platelet and fresh frozen plasma transfusions in the algorithm group. Another limitation of our study is that we used the maximum amplitude as our platelet function test. Newer point-of-care platelet function tests have recently become available, which may be better tests of platelet function and the need for desmopressin acetate or platelet transfusion therapy. A further limitation of our study is that no test for residual heparin was used beyond return of the ACT to within 10% of the preheparin value. Residual heparin could prolong the APTT after CPB. An algorithm in this study did not guide the erythrocyte transfusions. Therefore, the differences found in erythrocyte transfusions in the ICU should be viewed with caution. Furthermore, the control patients entered the ICU with a lower hemoglobin concentration than the algorithm patients, which may explain the difference in ICU erythrocyte transfusions.

There also were small differences in the incidence of preoperative warfarin use, preoperative platelet count, and the presence of other cardiac procedures, but these seem unlikely to account for the observed differences in blood component transfusions. Dietrich et al.  33found that preoperative phenprocoumon (a warfarin analog) therapy reduced bleeding after CPB. We performed a multivariate analysis to attempt to control for these factors and found there were still significantly fewer units of fresh frozen plasma and platelets transfused in the algorithm group. Furthermore, differences between groups are also unlikely because randomization of all 836 patients occurred when they enrolled in the study before surgery.

In conclusion, the cause of abnormal bleeding after CPB is multifactorial, and assessment of the coagulation system is needed to efficiently treat this bleeding. Intraoperative use of a coagulation test–based transfusion algorithm in cardiac surgery patients reduced allogeneic transfusions in the operating room. Because of large variances in transfusion practice among medical institutions, the overall impact of our transfusion algorithm must be weighed against local factors influencing decisions. A multicenter study will be needed to address this. If this algorithm is successful in other institutions, large reductions in both the amounts of blood products transfused and the institutional costs for the care of cardiac surgery patients may be seen.