Heart failure after cardiac surgery could lead to organ failure and death, therefore, adequate monitoring of hemodynamic state is very important in cardiac surgery patients [1,2]. Maintenance of adequate oxygen delivery is essential to keep organ function. Monitoring of mixed venous oxygen saturation (SvO2) with pulmonary artery catheter (PAC) is used for evaluation of oxygen delivery, and has been indicated as a prognostic predictor in cardiac surgery patients . However, catheterization of PAC is costly, and has the risk of various complications including pulmonary artery rupture, arrhythmia, pneumothorax, and air embolism . Measurement of central venous oxygen saturation (ScvO2) was suggested as a more accessible and simpler monitoring of global tissue oxygenation [5-10]. Maintaining ScvO2 more than 70% by an aggressive hemodynamic management protocol has been shown to reduce mortality in septic patients . As central venous catheterization is less invasive compared to the pulmonary artery catheterization, many trials were conducted to investigate that ScvO2 could be alternative to SvO2. Yet, in critically ill patients, the agreement of ScvO2 and SvO2 was not inconsistent and unsatisfactory [11-14].
ScvO2 measures the oxygen saturation of venous mixture from the upper body, while SvO2 determines the oxygen saturation of venous mixture from the whole (upper and lower) body including coronary sinus circulation . Therefore, in healthy subjects, ScvO2 value is slightly lower than SvO2 value . In critically ill patients, this situation is reversed. In the recent studies, the average value for ScvO2 was shown to be approximately 4.0% to 7.0% higher than that for SvO2 [6,16,17]. This is because changes in distribution of cardiac output occur in patients with shock state such as cardiac failure and sepsis.
The aim of this study is to test the hypothesis that in cardiac surgery patients, the increase of „SO2 during operation can predict postoperative complications. To assess the capability of „SO2 in predicting complications after cardiac surgery is an issue of great importance as it may lead to the new perioperative hemodynamic management protocol.
Subjects and Methods
Patients and Anesthesia
We obtained approval from the Ethics Committee of our hospital and informed consent from all enrolled patients. One hundred and two patients scheduled for cardiac surgery on cardiopulmonary bypass (CPB) were included in this study. Exclusion criteria were hemodialysis and intracardiac shunt. No premedication was administered. General anesthesia induction was performed using midazolam (0.05 to 1 mg/kg), propofol (1 to 2 mg/kg), fentanyl (2 to 3 g/kg), and rocuronium (0.6 to 1 mg/kg). After tracheal intubation, all patients were ventilated with a tidal volume of 8 to 10 ml/kg of ideal body weight. We controlled the frequency of mechanical ventilation to keep end-tidal carbon dioxide between 35 to 40 mmHg, and inspired oxygen concentration to maintain intraoperative PaO2 > 100mmHg. Anesthesia was maintained with sevoflurane (1.5 to 2.0%) and fentanyl (20 to 40 g/kg per case as total dose). During CPB, anesthesia was given with continuous propofol administration (2 to 3 mg/kg/hr). The depth of anesthesia was controlled to keep BIS (v. 4.0, Aspect Medical System, Natick, MA, USA) value from 40 to 60. Rocuronium (15 to 30 mg/hr) was administered to obtain muscle relaxation. After induction of general anesthesia, we inserted an arterial pressure line into the radial artery, and a central venous catheter (Presep catheter, Edwards Lifesciences, Irvine, CA, USA) and thermodilution pulmonary artery catheter (Edwards Lifesciences, Irvine, CA, USA) into the right internal jugular vein. The position of catheter was confirmed by transesophageal echocardiography and its pressure wave. After insertion of these catheters, measurement of ScvO2 and SvO2 was started. ScvO2 and SvO2 were measured before and after CPB.
Standard CPB procedures were performed in all patients. Standard flow rates of 2.6 L/min/m2 were utilized to maintain a mean arterial pressure from 50 to 80 mmHg. We maintained PaCO2 at 40mmHg or greater by alpha-stat management, and hematocrit above 22%. Mild hypothermia (32 OC at a rectal temperature), antegrade and retrograde crystalloid cardioplegia were used. Circulation arrest and cerebral perfusion were utilized in patients with thoracic aorta aneurysm. Circulation was arrested at a rectal temperature below 26 OC. Right-side cerebral perfusion via the right axillary artery was maintained 700 ml/min. Selective left-side cerebral perfusion to left common carotid artery and subclavian artery was maintained 250 ml/min. Mean arterial pressure during cerebral perfusion was managed from 40 to 50 mmHg at right radial artery.
Measurement of ScvO2 and SvO2 was discontinued in the operation room. After operation, all patients were transferred to the intensive care unit (ICU). Postoperative management was performed by the surgeons who were blinded to the intraoperative ScvO2 and SvO2 data. Criteria for discharge from the ICU were as follows; (a) respiratory stability defined as maintenance of SpO2 > 95% on < 5L supplemental O2; (b) hemodynamic stability defined as one and less intravenous inotropic drugs, removal of arterial and pulmonary artery catheters, and absence of unstable arrhythmias; (c) urine output > 1ml/kg/hr; (d) chest tube drainage < 10 ml/hr.
To compare the severity and incidence of postoperative complications, the incidence of major organ morbidity and mortality (MOMM) was considered as mentioned in the previous studies [18,19]. MOMM included prolonged ventilation more than 48 hours, renal failure requiring dialysis, stroke, reoperation, deep sternal infection, and death.
The ScvO2 and SvO2 values during operation were automatically collected. The average value of „SO2 (= ScvO2 - SvO2) for every 1 minute was calculated. To determine the threshold of „SO2 predicting postoperative complications in cardiac surgery, we made a receiver operating characteristic (ROC) analysis to evaluate prognostic performance of „SO2, ScvO2, and SvO2 with regard to prolonged ICU stay (> 3 days). We divided the patients into two groups according to this „SO2 threshold value made by ROC analysis (Group D; discrepancy of „SO2 [> the threshold value], Group N; no discrepancy of „SO2 [< the threshold value]). We compared these two groups at the point of perioperative complications. And, separate multivariate logistic regression models were utilized to investigate the independent effects of perioperative variables on the risk of developing prolonged ventilation (> 24 hr) and prolonged ICU stay (> 3 days).
All results were expressed as mean and standard deviation (SD) unless otherwise indicated. Statistical analysis was performed with SigmaPlot 11.2 (Systat Software Inc., San Jose, CA, USA). We used the Student's t-test and Mann-Whitney U test to compare the demographic data. For all analyses, a P value <0.05 was considered as significant.
One hundred and two patients undergoing cardiac surgery were enrolled. Perioperative characteristics in studied patients are shown in Table 1. The average value of minimum ScvO2, minimum SvO2, and maximum „SO2 was 68.1%, 69.2%, and 10.7%, respectively. At first, a ROC analysis was performed to evaluate prognostic performance of „SO2, ScvO2, and SvO2 with regard to prolonged ICU stay (> 3 days) (Figure 1). The area under the ROC curve was 0.745 for „SO2, which was significantly different from those of ScvO2 and SvO2 (p < 0.05) (ScvO2; 0.584, SvO2; 0.598). The optimal threshold value for „SO2 to predict prolonged ICU stay (> 3days) was 12% (sensitivity: 72.0%, specificity: 76.9%). We employed this threshold of 12% to divided the patients into 2 groups (Group D; intraoperative maximum „SO2 > 12% [n = 48], Group N; intraoperative maximum „SO2 < 12% [n = 54]).
Table 2 shows the preoperative data in both groups. There were significant differences between 2 groups in New York Heart Association (NYHA) classification and number of patients with renal insufficiency (serum creatinine > 1.2 mg/dl) (p < 0.05). Perioperative data in both groups are shown in Table 3. The number of patients with intraoperative use of intraaortic balloon pumping (IABP) was significantly higher in Group D patients. The ScvO2, and SvO2 values were significantly lower, and „SO2 values were significantly higher in Group D compared to those in Group N. The postoperative ICU duration, ventilation time and hospital stay were significantly longer in Group D patients than those in Group N patients (Table 4). As postoperative complications, the numbers of patients with postoperative use of IABP, delirium, respiratory failure requiring tracheotomy, and MOMM were significantly higher in Group D patients. We used multivariate logistic regression models to evaluate the independent effects of perioperative variables on the risk of developing prolonged ventilation (> 24 hr) and prolonged ICU stay (> 3 days), which are shown in Table 5 and 6. Discrepancy of intraoperative „SO2 was the independent risk factor of postoperative prolonged ventilation and ICU stay.
This is the first published study to investigate the correlation between the increase of „SO2 during operation and postoperative complications in cardiac surgery patients. We found that „SO2 was a better predictor than individual ScvO2 and SvO2 with regard to prolonged ICU stay, and the optimal threshold value for „SO2 was 12% (sensitivity: 72.0%, specificity 76.9%). This means that we can predict prolonged ICU stay with high sensitivity of 72% in patients with an increase of „SO2 (> 12%) during cardiac surgery. The 2 groups divided by the „SO2 value of 12% during operation were compared. We have shown that the patients with higher „SO2 (> 12%) had longer postoperative ICU stay, ventilation time, and hospital stay. The number of patients with MOMM was significantly greater in patients with higher „SO2 (> 12%). The patients with greater „SO2 (> 12%) had more severe complications than those with smaller „SO2 (< 12%). These results of our study indicate that „SO2 is a good predictor of postoperative complications in cardiac surgery patients and its reliability is significantly better than those of individual ScvO2 and SvO2.
In healthy subjects, the ScvO2 value is lower than the SvO2 value. Kidneys receive a high proportion of cardiac output, but do not have much oxygen consumption, therefore, blood in the inferior vena cava has a higher oxygen content than that in the superior vena cava . This is the reason why the SvO2 value is greater than the ScvO2 value. Several studies were conducted to investigate the clinical applicability of substituting ScvO2 for SvO2 in critically ill patients. [12,13,20,21]. In cardiac surgery, it was reported that SvO2 monitoring should not be replaced with ScvO2 monitoring, and increased oxygen extraction ratio and hypocapnia were the major factors to worsen the agreement between ScvO2 and SvO2 [12,13]. In patients with low cardiac output after coronary artery surgery, ScvO2 could not be used as an alternative for SvO2 . Yazigi et al.  also reported that in these patients, the disagreement between ScvO2 and SvO2 was not corrected after volume loading and normalization of cardiac output. In liver transplantation surgery, ScvO2 and SvO2 showed a good agreement during the preanhepatic stage, whereas the agreement was poor after liver graft reperfusion .
This disagreement of ScvO2 and SvO2 in critically ill patients is caused by the relation changes between ScvO2 and SvO2 in periods of cardiovascular instability. Changes in distribution of cardiac output occur in patients with hemodynamic instability. In the periods of hemodynamic instability, blood flow to the abdominal circulation decreases, while that to the important organs such as heart and brain is maintained . This leads to the decrease of oxygen content in the inferior vena cava, thus, ScvO2 value becomes higher than SvO2 value. In recent studies, the average ScvO2 value was shown to be 4.0% to 7.0% greater than SvO2 value in critically ill patients [6,16,17]. In patients with septic shock, the bias between ScvO2 and SvO2 was 8.45%, and this difference correlated significantly to the noradrenaline infusion rate . In heart failure state, bias between ScvO2 and SvO2 was 6.9%, and this disagreement appeared to be more significant when SvO2 was below 70% . This indicates that in patients with severe heart failure, the difference between ScvO2 and SvO2 becomes greater, and which corresponds to our results. In the present study, the patients with greater „SO2 (> 12%) had prolonged ICU stay, ventilation time, hospital stay, and more severe postoperative complications. These results seem to be caused by heart failure after cardiac surgery. The patients with greater „SO2 were in the circulatory failure states, and this might lead to postoperative complications.
Mozina et al.  reported that there was a significant difference between ScvO2 and SvO2 (9.4 + 7.5%) in patients with severe heart failure. They also showed that the difference between ScvO2 and SvO2 correlated significantly with plasma lactate level. Gutierrez et al.  investigated whether „SO2 were associated with survival rate in 106 critically ill patients. They drew blood samples from the proximal (ScvO2) and distal (SvO2) port of PAC every 6 hours from the time of PAC insertion until its removal (30.9 +11.0 hours), and reported that more survivors had mean positive „SO2 than descendents (p < 0.01). This result seems to contradict our results, however, they did not perform continuous monitoring of ScvO2 and SvO2. Continuous monitoring of ScvO2 and SvO2 is a critical issue when we use „SO2 as a predictor of clinical implications. If they performed continuous monitoring of ScvO2 and SvO2, the results might differ.
This study has some limitations. First, we did not perform monitoring of ScvO2 and SvO2 in the postoperative period. The „SO2 value in the ICU may reflect the postoperative complications more accurately than that during operation. However, as invasive surgical procedures could disclose the potential risk of heart failure evident, intraoperative observation is of significance. Second, we included the patients undergoing circulatory arrest and cerebral perfusion. This may affect the postoperative complications such as stroke and infection disease. Even with those limitations, our results suggest that in cardiac surgery patients, we can predict postoperative complications by measuring „SO2 value during operation.
The discrepancy between ScvO2 and SvO2 during cardiac surgery is the independent risk factor of postoperative complications such as prolonged ICU stay and ventilation time. We can predict the severity and incidence of postoperative complications by measuring „SO2 value during cardiac surgery. The reliability of „SO2 to predict postoperative complications may be better than those of ScvO2 and SvO2. The ScvO2 value is recommended to use in early-goal directed therapy, however, in cardiac surgery patients, both ScvO2 and SvO2 values should be monitored. Additionally, further studies are needed to investigate a new perioperative hemodynamic management protocol including measurement of discrepancy of ScvO2 and SvO2 in cardiac surgery.