Monitoring of brain oxygen saturation (INVOS) in a protocol to direct blood transfusions during cardiac surgery: a prospective randomized clinical trial

Cardiac surgery is associated with significant use of blood products. Although anemia must be avoided in these operations [1, 2], data suggest that blood transfusions are associated with worse outcomes [3, 4], and restrictive transfusion practices (hemoglobin concentration maintained between 7.0 and 9.0 g/dL) reduce organ dysfunction and cost, without adversely affecting outcome [5, 6]. A recent randomized controlled trial (RCT) in cardiac surgery showed that a restrictive transfusion strategy targeting hematocrit 24% reduced RBC use without increasing complications or mortality [7]. Furthermore, the number of transfused RBC units was a risk factor for worse outcomes and mortality in regression analysis.

Several variables associated with total red cell mass, such as preoperative anemia, female gender and small body size, are predictors of transfusion in cardiac surgery [8‐10]. Hematocrit and hemoglobin values are markedly affected by positive fluid balance while on CPB without true loss of erythrocytes. Existing guidelines underline the importance of limiting hemodilution [11, 12].

Blood transfusions based only on hematocrit value seem unjustified in cardiac surgery. The lack of a transfusion “threshold” and data from studies evaluating restrictive blood transfusion strategies suggest that hematocrit alone does not optimally support decisions to transfuse.

Monitoring of brain tissue oxygen saturation with near-infrared spectroscopy (INVOS) is a method for evaluating the balance of brain oxygen supply vs. consumption. Published data suggest that INVOS values have positive correlation with arterial hemoglobin oxygen saturation, arterial pCO2 values and hemoglobin (or hematocrit) values and negative correlation with patient age [13, 14]. The effect of anesthetic depth on INVOS values has not been thoroughly evaluated, but data from a small RCT have shown higher INVOS values with deeper levels of sevoflurane or desflurane anesthesia [15].

Our experience while conducting earlier studies and in everyday clinical practice suggests that brain oxygen saturation with near-infrared spectroscopy (INVOS) monito ring may help limit RBC use during CPB. The present study was conducted to investigate the hypothesis that use of intra-operative INVOS monitoring can reduce blood transfusions in patients undergoing cardiac surgery under CPB with a restrictive parenteral fluid protocol and blood salvage techniques.

In total, 150 patients were enrolled, and there were no cases of missing data. Despite occasional difficulty with placing both BIS and INVOS sensors in group A, all sensors were successfully applied and produced BIS and INVOS data of adequate quality on all group A patients. The CONSORT flowchart showing the flow of patients through the study and data analysis is presented in Figure 1. Demographic, clinical and perioperative data for both groups are summarized in Table 1, and did not differ significantly between groups. INVOS data recorded in group A are summarized in Table 2. Hematocrit values throughout the operation, during ICU stay and until the day of discharge, and data on fluid balance are shown in Table 3. Fluid balance for the entire procedure was calculated as [(total intravenous fluids + pump prime + total cardioplegia + any other “ extra ” volume in the CPB machine) − (total urine production + hemofiltration volume + saver net filtration + residual CPB circuit volume)], varied from −950 to 2550 ml in group A and from −550 to 2500 ml in group B. All variables shown on Table 3 did not differ significantly between the two groups.

Transfusion data by group and comparison between the groups based on “intention to treat” are summarized in Table 4. In total, 254 units of RBCs were transfused in 150 patients. Intraoperatively, fewer group A (INVOS) patients received fewer RBC units, and the difference was statistically significant (p = 0.04). Overall, 51 of 75 patients (68%) received transfusions in the INVOS group vs. 63 of 75 patients (84%) in the control group, and this difference was highly significant (p = 0.021). Hematocrit indication was present in 20 (26.6%) group A patients (all during CPB). Transfusion criterion a (mean INVOS value less than 60%) was present in 37 patients and criterion b (INVOS ≥20% decrease) in 27 patients. Among these patients, both criteria were present in 20 patients, not necessarily the same as those showing hematocrit indication for transfusion. In total, 14 patients in this group were transfused receiving 18 units of RBCs (Table 4). Hematocrit indication was present in 21 (28.0%) patients in group B, and the proportion of patients who met the hematocrit indication for transfusion did not differ significantly between the two groups (p = 0.854). These patients were transfused along with 4 more (see below: protocol violations) receiving in total 40 units of RBCs.

There were no OR deaths in either group. Duration of mechanical ventilation, ICU length of stay (LOS) and hospital LOS did not differ significantly between groups (Table 1). Fifty-one patients (68.0%) in group A and 41 patients (54.7%) in group B stayed in the hospital for more than 9 days, but most of these patients did not exhibit any pre-defined complication. In group A complications included: a) one patient for MVR/CABG = significant postoperative hemorrhage and valve dysfunction, emergency re-operation, transfusion of 15 units of RBCs, death in ICU in the 3rd day, b) ventilation time >24 h in 9 patients (3 patients were re-intubated for low cardiac output syndrome, 3 patients showed difficult and delayed weaning from mechanical ventilation, 1 patient required dialysis, 1 patient developed neurologic deficit delaying extubation but with good course, 1 patient was re-intubated for cardiac arrest during removal of thoracic drains and was re-explored), c) MI in 2 patients, and d) arrhythmias requiring treatment in 3 patients. In group B complications were: a) one patient was re-intubated for arrhythmia, hemodynamic instability, increase of hepatic enzymes and finally he died, b) ventilation time >24 h in 6 patients (1 patient was re-intubated for pneumothorax and low cardiac output syndrome, 1 patient for MI and low cardiac output syndrome, 1 patient for re-exploration for bleeding, 3 patients showed difficult and delayed weaning from mechanical ventilation), c) MI in 2 patients, d) arrhythmia requiring treatment in 2 patients, e) hemodynamic instability with cognitive dysfunction in 1 patient and f) treatment for hypertension delaying discharge in 1 patient. The overall mortality in this study was 2 deaths in 150 patients.

In total, protocol violations were identified for 13 patients (8.7%). Five of them were noted in group A (6.6%): a) one patient received 15 units of RBCs and finally died, b) two patients undergoing CABG received transfusions even though they did not meet criteria for transfusion according to our protocol, c) one patient undergoing CABG was intraoperatively transfused after meeting all (INVOS and hematocrit) criteria, but received parenteral fluids to a positive balance of 2550 ml, c) one patient undergoing CABG met criteria for transfusion but fluid balance positive by over 1.5 L. The protocol was violated in 8 group B patients (10.6%): a) one patient undergoing CABG was transfused under hematocrit indication but had postoperative fluid balance > +1.5 L; b) three patients undergoing CABG and one patient undergoing MVR were transfused without meeting the hematocrit indication (and all of them with positive fluid balance > 1.5 L) and c) three patients (one undergoing AVR/CABG and two undergoing CABG) did not receive transfusions, but had positive fluid balance >1.5 L.

After excluding cases of protocol violation, statistical analysis “per protocol” showed that, compared to the control group, significantly fewer group A patients received any blood transfusion, both intraoperatively (11 of 70 patients in group A vs. 20 of 67 patients in the control group, p = 0.048) and during the entire hospitalization (46 of 70 patients in group A vs. 55 of 67 patients in the control group, p = 0.029). Similarly, compared to the control group, the total number of RBCs transfused per patient was significantly lower in the INVOS group intraoperatively (0.20 ± 0.50 units per patient in the INVOS group vs. 0.52 ± 0.88 units per patient in the control group, p = 0.008), and during the entire hospitalization (1.31 ± 1.20 units per patient in the INVOS group vs. 1.82 ± 1.46 units per patient in the control group, p = 0.024). Of note, because the number of patients transfused and the number of RBC units transfused per patient did not differ between groups during ICU stay, the overall lower RBC use in the INVOS group seems to be the result of reduced intraoperative RBC use, while the INVOS-based protocol was in effect. Results of a per protocol analysis of our transfusion data (after excluding cases of protocol violation) are presented in Table 5.

Transfusion of RBC and other blood components in common in cardiac surgery, due to several reasons, including pre-existing anemia, intraoperative hemodilution and intraoperative or postoperative blood loss, even in otherwise uncomplicated surgery. However, even though transfusions are clearly necessary in many cases, blood product use has been associated with increased morbidity [4] and long-term mortality [3] in cardiac surgery, and therefore attempts to reduce blood transfusions may contribute to reduced complications and improved outcomes. Consequently, several measures have been evaluated in an attempt to reduce blood transfusions during cardiac surgery [12], including tolerating lower hemoglobin values [5] and restricting parenteral fluids in order to avoid hemodilution [16]. Studies evaluating the effectiveness of restrictive fluid administration strategies in mitigating the precipitous intraoperative hematocrit drop, rather than restrictive blood transfusion strategies, are scarce. We recently reported that, compared to liberal fluid administration, a restrictive parenteral fluid protocol significantly reduced intraoperative red blood cell (RBC) transfusions [17]. We also showed in a later RCT that restrictive fluid strategies may be more beneficial in patients prone to transfusion because of low preoperative hematocrit, female sex, or small BSA [16]. In both studies, re-infusion of washed blood from the thoracic cavities was used to reduce the need for transfusion. However, we faced difficulty defining a transfusion “threshold” in these studies, because existing reports raise concerns regarding safety when tolerating low hematocrit values [18, 19].

Because INVOS reflects the balance of site-specific oxygen delivery vs. consumption, it is plausible to consider INVOS a reasonable monitor for adequacy of oxygen tissue delivery. Furthermore, because there is a positive correlation between INVOS and hemoglobin concentration [13], it is reasonable to interpret low INVOS values as evidence that hemoglobin or hematocrit values are not sufficient for maintaining adequate oxygen delivery to tissues, and therefore the patient needs to be transfused. Similarly, adequate INVOS values could be considered as indicator of adequate oxygen delivery, and therefore as reassurance that transfusion of red blood cells is not needed. However, data on the use of INVOS as tool for decision making with regards to blood transfusion during cardiac surgery are, to our knowledge, very limited, and we believe there is a need for rigorous assessment of this technology in cardiac surgery and perhaps in other types of major surgery.

The main finding of this prospective, randomized study was the significantly reduced intraoperative and overall need for transfusion of RBCs in patients undergoing elective cardiac surgery, with use of a hematocrit and INVOS-supported algorithm to guide the decision to transfuse, and the observed differences between the two groups were significant both in the “intention to treat”. Significant findings derived from “intention to treat” analysis are generally preferred, because results based on the “intention to treat” principle better reflect “real life” conditions, whereby errors in judgment and deviations from established protocols invariably occur. We believe that, our findings are worth reporting, because they suggest that, under optimal conditions (“per protocol analysis”) and in real-life conditions (“intention to treat” analysis), the use of INVOS to assist with decision-making regarding blood transfusions could be of value, and therefore deserves further study. In addition, we believe it is worth noting that the overall lower RBC use in group A (the INVOS group) was the result of reduced RBC use intraoperatively (while the INVOS-based protocol was in effect). Therefore, it may be reasonable to speculate that the observed overall differences between the two groups could be greater if the INVOS-protocol were applied in the ICU as well. Clearly, this is just a hypothesis, but we think this hypothesis deserves to be evaluated in a formal study.

Strengths of our study include study design (prospective, randomized, blinded), the inclusion of a well defined patient population, with clearly defined criteria for transfusion. In addition, when protocol was properly followed, reduction of RBC use was highly significant, and the observed difference between groups was clinically meaningful. Furthermore, the use of a restrictive intravenous fluid administration policy on all patients probably resulted in overall reduction of transfusion requirements in both groups, due to avoidance of hemodilution before and during CPB. Consequently, it is plausible that the observed differences between the groups, and therefore the perceived benefit from using INVOS as tool to guide transfusion decisions could be more pronounced in settings where fluid restriction is not used, and therefore overall transfusion requirements are higher.

Disadvantages of the study include the significant number of cases with protocol violations; this is probably the reason why data analysis based on “intention to treat” did now show a significant difference with regards to overall number of RBC units transfused per patient between the INVOS and the control group. Consequently, our results also demonstrate that it is difficult to conduct clinical studies in the complex operating room and ICU environment utilized for cardiac surgery, and, despite efforts to avoid them, study protocol violations can still occur. Therefore, researchers and study coordinators need to pay particular attention and closely monitor such studies in an attempt to minimize protocol violations and therefore generate more valid data.

This prospective randomized clinical study suggests that the use of cerebral oximetry (INVOS) as part of an algorithm to guide RBC transfusions can result in significant reduction of RBC use in patients undergoing elective cardiac surgery, when the established protocol for fluid restriction and decision to transfuse is properly followed. However, the observed benefit is no longer significant when protocol violations are included in the analysis (“intention to treat” analysis). We suggest that, based on these results, INVOS could be a useful tool for monitoring patients during cardiac surgery, but data from well designed clinical trials with rigorous attention to study protocol, in an attempt to minimize protocol violations are needed to better assess the validity of our findings.