Restrictive Transfusion Strategy after Cardiac Surgery: Role of Central Venous Oxygen Saturation Trigger: A Randomized Controlled Trial

The study was designed as a single-center, single-blinded, randomized controlled trial. It was approved by the institutional review board and the national drug safety agency and registered at ClinicalTrails.gov (NCT 02761564; principal investigator: Norddine Zeroual, M.D.; May 4, 2016). The full trial protocol can be obtained upon request. Eligible patients signed an informed consent before surgery and agreed that they could be included in the study should anemia occur in the ICU.

Adult patients (older than 18 yr) scheduled for elective or urgent (meaning during the index hospitalization of a coronary syndrome or an endocarditis) on-pump cardiac surgery were screened (eligibility assessment) before surgery (inclusion/exclusion criteria reported in table 1). After surgery, whenever a postoperative anemia (hemoglobin less than 9 g/dl) occurred, patients were randomly assigned to the study groups (see below, Transfusion Management: In the ICU) provided they fulfilled three conditions: (1) the central venous catheter inserted in the superior vena cava routinely before surgery (checked by chest x-ray) was functional; (2) no early active bleeding; and (3) no ongoing transfusion. Randomization was postponed if the patient had significant bleeding within the first postoperative hours, as assessed by a blood loss of more than 2 ml · kg^–1 · h^–1, or if the patient was receiving a transfusion decided before ICU admission. When the active bleeding decreased significantly, or after the ongoing transfusion, anemia was re-checked and the patient was eventually randomized. To study anemia in stable condition, besides delaying the randomization of patients with sustained early bleeding, those among them who cumulated more than 1,000 ml or those requiring a transfusion of more than 4 erythrocyte units within the first 12 postoperative hours (severe bleeding class according to the Universal Definition of Perioperative Bleeding)¹⁸  were excluded from the study.

The first occurrence of hemoglobin less than 9 g/dl in the ICU fixed the group assignment according to the randomization. However, the design of the study anticipated including all transfusions in the final count, i.e. until postoperative day 28 or hospital discharge. The study was conducted in the cardiac surgery ICU of a tertiary university hospital and in the surgical ward until postoperative 28 or hospital discharge.

The usual cardiac treatment was maintained until the day before surgery except for antiplatelets or oral anticoagulants, which were stopped as recommended in guidelines. Perioperative management (anesthesia, cardiopulmonary bypass [CPB], and cardioplegia) is standardized for all patients. CPB was set up with a membrane oxygenator primed with 1.5 l of crystalloid (Isofundin; B.Braun, Germany) and with body temperature maintained at 34° to 36°C. After aortic clamping, a hyperkalemic blood cardioplegic solution was infused into the aortic root of the aorta until myocardial arrest. A nonpulsatile pump flow rate was maintained between 2.2 and 2.6 l · min^–1 · m^–2, provided the mixed Svo₂ monitored on the pump machine exceeded 60%. A crystalloid solution (balanced crystalloid solution) could be added to maintain pump flow, and blood products were used to maintain the hemoglobin concentration at greater than 7 g/dl during CPB and greater than 8 g/dl after CPB. Hypotension (mean arterial pressure [MAP] less than or equal to 60 mmHg) was treated with a continuous intravenous administration of norepinephrine in order to maintain MAP greater than 60 mmHg. After completion of the surgical procedure, patients were weaned from CPB when a rectal temperature of 36°C had been reached.

In the ICU, inotropic support (dobutamine) was used to obtain a cardiac index greater than or equal to 2.2 l · min^–1 · m^–2 output as assessed by thermodilution (pulmonary catheter) or echocardiography (aortic flow measured by pulsed Doppler time velocity integral) with normal left ventricular filling pressure. Hypotension (MAP less than or equal to 60 mmHg) without cardiogenic shock features (cardiac index greater than 2.2 l · min^–1 · m^–2) was treated with a continuous intravenous administration of norepinephrine in order to restore MAP to greater than 60 mmHg; repeated boluses of morphine were used to keep patients pain-free. Weaning from the ventilator was started during the emergence of anesthesia and extubation was performed when stable hemodynamics and normothermia had been maintained for greater than or equal to 1 h. Oxygen supply was provided to maintain oxygen saturation measured by pulse oximetry at greater than or equal to 94%.

A patient blood management protocol was implemented in our institution in 2016. It specified how to manage anemia, how to optimize erythrocyte mass, and how to minimize blood loss. Besides the routine use of cell savage, perioperative tranexamic acid administration, and anemia medical treatment (iron supplementation, erythropoietin) when appropriate, the patient blood management protocol established hemoglobin thresholds for transfusion, either during CPB (less than or equal to 7 g/dl) or in postoperative care, in the ICU, and in the surgical ward (less than 9 g/dl). Hemoglobin was measured (1) at arrival in the ICU and (2) every 12 h during the first 2 days after surgery, then (3) once daily in the ICU and (4) once a week and at the physician’s discretion in the surgery ward, as well as (5) after any transfusion.

Patients assigned to the control group were transfused if hemoglobin was less than 9 g/dl. Transfusion was repeated whenever hemoglobin was less than 9 g/dl.

Patients assigned to the Svo₂ group were transfused if hemoglobin was less than 9 g/dl and central Svo₂ less than or equal to 65%. When hemoglobin was less than 9 g/dl and central Svo₂ greater than 65%, no transfusion was administered. Transfusion was repeated whenever hemoglobin concentration was less than 9 g/dl and central Svo₂ less than or equal to 65%. The central Svo₂ less than or equal to 65% threshold was chosen based on a previous observational study, which showed that 85% of anemic patients with central Svo₂ less than or equal to 65% responded to transfusion by an increase in central Svo₂ of at least 5%.¹⁷  Central Svo₂ was measured on a blood sample obtained from the distal lumen of the central venous catheter. Oximetry was used for measurement with a point-of-care of gas analysis located in the ICU (Gem 4000; Instrumentation Laboratory, USA).

Physicians were permitted to deviate from the hemoglobin/Svo₂ criterion when deciding whether to transfuse in the case of critical clinical conditions including: (1) severe sepsis; (2) septic shock; (3) hemodynamic instability (defined either by the introduction of inotropic or vasopressor treatment, or by a significant change in the dosage [increase or decrease of more than 20%]); or (4) an episode of severe pulmonary failure (defined by partial pressure of oxygen in the arterial blood on the fraction of the inspired oxygen ratio as less than 200). If physicians decided to transfuse Svo₂ patients when central Svo₂ was greater than 65%, they had to specifically document their reason.

Other blood components, fresh frozen plasma, platelets, fibrinogen, or prothrombin complex concentrates were also administered according to our institutional patient blood management protocol. In brief, they were administered in order to maintain the prothrombin time ratio at greater than 50 to 70%, the platelets at above 50 to 100 gigas, and the fibrinogen plasma concentration at greater than 1.5 to 2 g/l, (the lower range limits without active bleeding, and the upper range limits in the case of active bleeding).

Until hospital discharge or postoperative 28 (study discharge), transfusion was managed according to the patient blood management protocol in both groups, meaning that transfusion was considered when hemoglobin was less than 9 g/dl with no further central Svo₂ measurements in the Svo₂ group.

The primary endpoint was the proportion of patients transfused during the ICU stay. Secondary endpoints consisted of transfusion and postoperative outcomes. For transfusion, we collected the total number of erythrocyte units transfused in the ICU, the proportion of patients transfused, and the total number of units transfused at study discharge. Postoperative outcomes included the incidence of mortality and postoperative complications at study discharge, along with the length of the ICU and hospital stays.

Postoperative complications were defined as infection (respiratory, septicemia), ischemic events (myocardial infarction, stroke, mesenteric ischemia), organ failure (renal dysfunction according to Kidney Disease: Improving Global Outcomes classification), or liver dysfunction (if plasma hepatic enzymes or bilirubin exceeded twofold the standard value). A composite outcome included mortality and/or any of the aforementioned postoperative complications. The Simplified Acute Physiology Score was collected within the first 24 h in the ICU to assess disease severity on ICU admission. The bleeding volume was quantified by the amount of blood collected by surgical drains at postoperative day 1.

According to a previous study,¹⁶  we could expect a reduction of at least 20% in the transfusion incidence of the Svo₂ group, compared to 100% in the control group. A sample size of 100, with 50 patients in each group, would be sufficient for a superiority hypothesis to have 90% power to detect the 20% difference with a two-sided test at a 5% significance level. We used the StudySize 2.0 tool that gives N = 45.25 for each group and 50 patients per group after an increase of 8%.

Before surgery, the research assistant evaluated the patients’ eligibility, and the investigators obtained their informed consent. The participants were then enrolled only if they fulfilled the inclusion criteria in the ICU. Patients screened before surgery underwent balanced randomization 1:1 and blocks of random size to either the control group or the Svo₂ group using a central internet-based randomization system (Cs Random online software; Ennov Clinical, France). The data manager of the study set up the central internet-based randomization system according to the protocol. The patient was blinded from his/her randomization arm. Care providers were aware of the group assignment.

The primary and secondary outcomes of all randomized patients were analyzed on an intention-to-treat basis. Baseline characteristics were reported using the mean and SD for normally distributed data (Kolmogorov−Smirnov test), and median with interquartile range [25 to 75] for other cases. Numbers and percentage were reported for qualitative cases.

The main endpoint (proportion of patients transfused during the ICU stay) was compared between the two groups using a chi-square test. The effect size was estimated with the odds ratio and its 95% CI.

For the secondary endpoints, the total number of erythrocyte units transfused in the ICU and at study discharge, and the ICU length of stays were compared between the two strategies. The Student t or Mann−Whitney tests were used according to the distribution of continuous data. The proportion of patients transfused, the incidence of mortality, and the postoperative complications at study discharge were all compared using Fisher exact test or expansion of Fisher test by A.W. Ghent.¹⁹  Firth logistic regression²⁰  was used to avoid any problems of convergence and to mitigate the bias of maximum likelihood estimates. When the model did not converge, exact logistic regression²¹  was applied.

All tests were two-sided. A probability value of less than 0.05 was considered to be significant for all statistical tests. Statistical analysis was performed using SAS Enterprise Guide, version 7.3 (SAS Institute, USA).

Between November 2016 and August 2018, 484 patients were screened. One hundred and thirty-seven (28.3%) had anemia (hemoglobin less than 9 g/dl) in the ICU, of whom 37 (7.6%) met exclusion criterion. The remaining 100 patients were included in the study, randomized (50 in each group) and were analyzed (fig. 1). During follow-up, four patients died: two from each group. There were no missing data (except those of the deceased patients), nor adverse events related to the protocol.

All the patients were analyzed and followed up to study discharge. The trial was conducted in accordance with the original protocol and was stopped when the target enrollment was obtained.

The baseline characteristics are reported in table 2. Most patients were men (63%), the mean age was 68.8 ± 8.8 yr, 27% underwent urgent surgery, and 40% received erythrocyte transfusion during surgery.

In the ICU, we observed 77 anemia episodes in the control group and 82 in the Svo₂ group. Sixty-two (62%) patients had only one episode of anemia and 38 (38%) had repeated occurrences with no significant difference between the control group (n = 17) and the Svo₂ group (n = 21; odds ratio, 1.39 [95% CI, 0.62 to 3.14]; P = 0.411; fig. 2). Among the 21 Svo₂ patients who had repeated episodes of anemia: (1) six (29%) had only anemia episodes with central Svo₂ greater than 65% each time and were not transfused; (2) five (24%) had only anemia episodes with central Svo₂ less than or equal to 65% each time and were transfused at each episode; and (3) 10 (48%) had both—anemia episodes with central Svo₂ less than or equal to 65% and anemia episodes with Svo₂ greater than 65%—and were transfused only when central Svo₂ was less than or equal to 65% (fig. 3).

In the ICU, there was no significant difference between the two groups in mean hemoglobin concentration (table 3). Similarly, in the Svo₂ group, hemoglobin concentration was not significantly different between patients with central Svo₂ less than or equal to 65% and patients with central Svo₂ greater than 65% (7.9 ± 0.9 g/dl vs. 8.1 ± 1.2, respectively; P = 0.370). However, central Svo₂ was significantly different (58.7 ± 7.0% vs. 72.0 ± 8.5%, respectively; P < 0.001).

The restrictive central Svo₂–guided protocol resulted in a 32% reduction of the proportion of patients transfused in the ICU. Thirty-four (68%) patients in the Svo₂ group were transfused in the ICU, compared to all patients (n = 50) in the control group in agreement with our standard protocol (odds ratio, 0.03 [95% CI, 0 to 0.15]; P < 0.001).

In the Svo₂ group, 16 patients were not transfused because central Svo₂ was always greater than 65% during their anemia episodes in the ICU. Their mean Hb was at 8.2 ± 0.8 g/dl (vs. 8.9 ± 0.8 g/dl in control; P = 0.023) at ICU discharge. Among the 34 transfused patients, 31 had central Svo₂ less than or equal to 65% and 3 had central Svo₂ greater than 65%. These three study protocol violations, which resulted in transfusion despite central Svo₂ greater than 65%, were motivated by the weaning of vasopressor support in a context of low Hb concentration (6.5, 6.8, and 7.1 g/dl, respectively).

In the ICU, 94 erythrocyte units (per patient median, 1 [interquartile range, 1 to 2]) were transfused in the control group and 65 (per patient median, 1 [interquartile range, 1 to 2]) in the Svo₂ group (table 3; 31% reduction compared to the control group). Between ICU discharge and study discharge, 14 patients were transfused in the Svo₂ group, including six who had already been transfused in the ICU, while 10 control patients were retransfused (P = 0.315; table 3).

At study discharge, eight patients of the Svo₂ group remained nontransfused (table 3). The cumulative count of erythrocyte units at study discharge was 126 in the control group and 96 in the Svo₂ group (a 24% reduction compared to the control group; table 3).

The hemoglobin concentration at study discharge was not significantly different between the two groups (table 2). For the eight Svo₂ nontransfused patients, hemoglobin was 9.1 ± 0.8 g/dl (vs. 10.0 ± 1.0 g/dl in control; P = 0.035) at study discharge.

At ICU admission, the Simplified Acute Physiology Score was 45 (interquartile range, 37 to 52) for the control group versus 46 (interquartile range, 39 to 56) for the Svo₂ group (P = 0.370). The bleeding volume at postoperative day 1 reached 665 ± 395 ml in the control group and 658 ± 362 ml in the Svo₂ group (P = 0.920).

Postoperative complications and mortality incidences are reported in table 4. Incidence of mortality, ischemic and septic complications, organ dysfunction, or composite outcome were not statistically different between the two groups.

Four patients in each group needed early reoperation for bleeding (8%). Intravenous iron supplementation (ferric hydroxide) was administered in 29 (60%) control versus 35 (74%) Svo₂ patients (odds ratio, 1.88; 95% CI, 0.79 to 4.5; P = 0.144). Erythropoietin (Epoetin beta) was administered in eight (17%) control versus four (8.5%) Svo₂ patients (odds ratio, 0.48; 95% CI, 0.14 to 1.65; P = 0.216).

This prospective randomized study shows that in anemic patients having undergone cardiac surgery, restrictive transfusion according to central Svo₂ less than or equal to 65% allowed a significant reduction in transfusion incidence in the ICU. The benefit of the central Svo₂–guided restrictive strategy was partially maintained (about half) after the ICU stay.

The main objective was to evaluate the impact of the central Svo₂–guided strategy on the incidence of transfusion. The originality of the study design was to compare a control group with a transfusion incidence of 100% to the study group, for which we expected a transfusion incidence less than or equal to 80%. Considering the whole cohort (484 screened patients), 137 (28.3%) were candidates for transfusion according to our standard of care. This proportion is in quite a low range compared to previous studies, possibly because of the implementation of the patient blood management protocol in our routine practice.^2,3  Nevertheless, the study demonstrated that 45% of the anemia episodes in the ICU were associated with central Svo₂ greater than 65%; the central Svo₂–guided restrictive strategy therefore resulted in a 32% reduction of transfused patients.

The study provides new information on the restrictive transfusion strategy. First, central Svo₂ appears to be a good complementary parameter to hemoglobin for helping in the transfusion decision. The hemoglobin concentration at the time of evaluating the need for transfusion was rather low (8.0 g/dl) and did not differ between groups, not even between patients in the Svo₂ group, whether central Svo₂ was greater than or less than or equal to 65%. Thus, central Svo₂ was actually the parameter that determined the decision to transfuse or not. Central Svo₂, like the mixed Svo₂, reflects the balance between the amount of oxygen delivered to the organs (i.e., oxygen delivery) and the amount of oxygen that the body takes up and utilizes (i.e., tissue oxygen consumption). To measure mixed Svo₂, a pulmonary artery catheter is required to get mixed blood from the whole body. Central Svo₂ requires only a blood sample collected from the central venous line and reflects the venous oxygen saturation of the superior vena cava territory. The measurement of central Svo₂ is therefore less influenced by the oxygen saturation of the inferior vena cava compartment and coronary sinus than that of mixed Svo₂. Nevertheless, there is a satisfactory correlation between central and mixed Svo₂, despite a lack of exact numerical equivalence.^22,23  For a given tissue oxygen consumption, central Svo₂ varies with the arterial oxygen content of blood, which depends on arterial oxygen saturation (Sao₂) and hemoglobin, and with cardiac output. In the conditions of the study, we can assume that Sao₂ and tissue oxygen consumption were stable, and that a low central Svo₂ meant that cardiac output did not compensate the low hemoglobin. Hemodynamic data were similar in both groups, but no measurement of cardiac output was available. Nevertheless, transfusion seems to be the most appropriate treatment in these cases for the following reasons: (1) increasing cardiac output with inotrope would have increased heart rate and myocardial oxygen consumption, whereas arterial oxygen content was low; therefore, patients would have been exposed to a risk of myocardial ischemia; and (2) even though increasing cardiac output by increasing ventricle preload is the best option, fluid loading without red cells would have worsened anemia, and transfusion would have been mandatory anyway. Incidentally, we cannot exclude that transfusion may have also optimized preload. Conversely, central Svo₂ greater than 65% suggested a good adaptation of cardiac output to match oxygen delivery with tissue oxygen consumption, despite the low hemoglobin. Therefore, central Svo₂ can help to identify, beyond “macrohemodynamic” parameters, anemic patients who may not need urgent transfusion, and who can instead receive medical treatment for anemia. As a consequence of this strategy, more than 30% of the patients avoided transfusion in the ICU.

Second, the study results go beyond the incidence of transfusion after cardiac surgery. The analysis of the repeated anemia episodes for each patient shows that patients can alternate anemia episodes with low or high central Svo₂ during the ICU stay. Indeed, patients had up to five anemia episodes, of which a small majority (55%) were associated with central Svo₂ less than or equal to 65%. Among the 21 (42%) Svo₂ patients who had several anemia episodes, almost half (n = 10) had a variable alternation of central Svo₂ less than or equal to 65% and greater than 65%. Moreover, there were similar numbers of repeated anemia episodes in both groups, suggesting that the 32% reduction of transfusion incidence in the Svo₂ group induced only a modest increase in anemia episodes (+6.5%). This result suggests that central Svo₂ may allow for adjustment of transfusion at each anemia episode, favoring individualization of transfusion.

Third, despite the change of transfusion strategy in the Svo₂ group in the surgical ward, we still observed a 16% reduction in the proportion of patients exposed to transfusion, and a 24% reduction in the erythrocyte units transfused at study discharge. In the surgical ward, the loss-of-central-Svo₂ trigger for guiding transfusion brings patient blood management back to the standard of care. However, the effect of the central Svo₂–guided restrictive strategy in the ICU, which spared 29 erythrocyte units, was still preserved in the surgical ward. Of the nontransfused patients in the ICU, 50% remained not transfused with a total of 30 erythrocyte units economized at discharge

. Adding to the fact that the hemoglobin concentration at study discharge was not statistically different to that of the control group, these observations suggest that the alternative treatment of anemia was effective enough to obtain an hemoglobin of around 10 g/dl in the Svo₂ group.

Owing to these observations, we believe that central Svo₂ may be a valuable parameter for adjusting the hemoglobin trigger for transfusion, and that it could be incorporated into the patient blood management strategy. However, further studies are needed to assess whether these treatments can be used more extensively after cardiac surgery, in the aim to reduce transfusion without long-term deleterious effects.

Patient inclusion based on postoperative anemia resulted in selecting the transfused population of our center, a subgroup of patients known to be at higher risk of postoperative complication.^5,7,8  The study was underpowered for safety considerations, but we certainly observed a high incidence of kidney injury (65%) and long ICU stays, which was similar in both groups. The outcome, defined as a composite of ischemic or infection complications and/or mortality, was also similar in both groups, and was close to the results of a previous study with a larger cohort.¹⁵  However, the exact role of restrictive transfusion or other postoperative complications like heart failure, acute kidney injury, or sepsis in a worsening outcome is difficult to establish as mechanisms may be entangled. Restrictive transfusion is noninferior to liberal strategies in short- and long-term outcomes,^12,13  but avoiding transfusion may delay anemia correction.^12,15  If transfusion is to be avoided when unnecessary, keeping patients anemic may expose them to complications.^24,25  A more aggressive treatment of anemia with iron supplementation and erythropoietin in per operative care has already been proposed and has shown encouraging perspectives even with short-term treatment.^26–28  We can assume that the extensive use of iron supplementation (60% of patients in the control group and 74% in the Svo₂ group) had contributed to restoring hemoglobin concentration within a few days.^27,28  However, studies are needed to assess whether these treatments can be used more intensively after cardiac surgery to reduce further transfusion and postoperative complications.

This study has several limits: it is a single-center study. Given that we used the patient blood management protocol as recommended by scientific societies, the results may be extended to centers with a similar practice. However, we cannot exclude bias in relation to the lack of blinding despite the fact that transfusion management was guided by predefined rules, and that very few protocol deviations were observed. Central Svo₂ was used as an adjustment variable to hemoglobin concentration in the transfusion decision. This may appear as a good indication of metabolic equilibrium, but depends on several factors beyond hemoglobin, such as cardiac output, Sao₂, and arterial oxygen content, which should be stable to allow correct interpretation during anemia. Moreover, circumstances where oxygen extraction is impaired, such as during sepsis or systemic inflammation, may render Svo₂ unreliable.

Though the number of included patients is rather small, it is noteworthy that there were no missing data. We deplored only two deaths in each group and there were no adverse events in relation to the protocol, which contributes to the strength of the study, beyond the randomization.

The study was not designed to assess long-term outcome, as it was not powered for this consideration. Therefore, although the complication incidence was in the range observed previously in a larger study on restrictive transfusion strategy,¹⁵  further studies are needed to confirm these preliminary results.

In conclusion, in this randomized controlled study, a restrictive transfusion strategy adjusted with central Svo₂ in complement to the hemoglobin concentration may allow a significant reduction in the proportion of patients with erythrocyte transfusion in comparison to a standard restrictive transfusion.

Support was provided solely from University Hospital of Montpellier, Montpellier, France.

The authors declare no competing interests.

Full protocol available at: p-colson@chu-montpellier.fr. Raw data available at: p-colson@chu-montpellier.fr.