The current study was designed to determine the relation between preoperative cerebral oxygen saturation (Sco2), variables of cardiopulmonary function, mortality, and morbidity in a heterogeneous cohort of cardiac surgery patients.
In this study, 1,178 consecutive patients scheduled for on-pump surgery were prospectively studied. Preoperative Sco2, demographics, N-terminal pro-B-type natriuretic peptide, high-sensitive troponin T, clinical outcomes, and 30-day and 1-yr mortality were recorded.
Median additive EuroSCORE was 5 (range: 0-19). Thirty-day and 1-yr mortality and major morbidity (at least two major complications and/or a high-dependency unit stay of at least 10 days) were 3.5%, 7.7%, and 13.3%, respectively. Median minimal preoperative oxygen supplemented Sco2 (Sco2min-ox) was 64% (range: 15-92%). Sco2min-ox was correlated (all: P value <0.0001) with N-terminal pro-B-type natriuretic peptide (ρ: -0.35), high-sensitive troponin T (ρ: -0.28), hematocrit (ρ: 0.34), glomerular filtration rate (ρ: 0.19), EuroSCORE (τ: 0.20), and left ventricular ejection fraction class (τ: 0.12). Thirty-day nonsurvivors had a lower Sco2min-ox than survivors (median 58% [95% CI, 50.7-62%] vs. 64% [95% CI, 64-65%]; P < 0.0001). Receiver-operating curve analysis of Sco2min-ox and 30-day mortality revealed an area-under-the-curve of 0.71 (95% CI, 0.68-0.73%; P < 0.0001) in the total cohort and an area-under-the-curve of 0.77 (95% CI, 0.69-0.86%; P < 0.0001) in patients with a EuroSCORE more than 10. Logistic regression based on different EuroSCORE categories (0-2; 3-5, 6-10, >10), Sco2min-ox, and duration of cardiopulmonary bypass showed that a Sco2min-ox equal or less than 50% is an independent risk factor for 30-day and 1-yr mortality.
Preoperative Sco2 levels are reflective of the severity of cardiopulmonary dysfunction, associated with short- and long-term mortality and morbidity, and may add to preoperative risk stratification in patients undergoing cardiac surgery.
What We Know about This Topic
Intraoperative cerebral oxygen saturation (Sco2) monitoring has been used to assess the adequacy of cerebral oxygen delivery to demand.
What This Article Tells Us That Is New
Preoperative Sco2concentrations are reflective of baseline severity of cardiopulmonary dysfunction, associated with short- and long-term mortality and morbidity, and may add to preoperative risk stratification in patients undergoing cardiac surgery.
CEREBRAL oxygen saturation (Sco2) monitoring by near-infrared spectroscopy is increasingly used for assessing the adequacy of cerebral oxygen delivery to demand in patients undergoing cardiac and noncardiac surgery.1–3A relative decrease of intraoperative Sco2to less than 80% of the preoperative baseline or to absolute levels lower than 50% have been associated with postoperative cognitive dysfunction,1,4a higher rate of stroke or coma,5increased postoperative morbidity from other nonneurological causes,6and a prolonged stay in the high-dependency unit or the hospital.7Sparse data are available about the relation between preoperative Sco2levels and clinical outcomes in cardiac surgery patients. Murkin et al. have shown that patients presenting with major organ morbiditiy and mortality after coronary artery bypass grafting (CABG) had lower baseline and nadir Sco2concentrations than those that did not.6Based on these findings and our own clinical observations that cardiac surgery patients presenting with decreased preoperative Sco2(i.e. , lower than 50% absolute) often developed complications and needed prolonged therapy in the intensive care unit, we hypothesized that preoperative Sco2is reflective of the global cardiopulmonary function and may be used for risk stratification. The current study was designed to explore the association of preoperative Sco2concentrations and established risk factors in patients scheduled for on-pump cardiac surgery and to determine the relation between these variables and clinical outcomes (i.e. , mortality and morbidity).
Materials and Methods
The study was designed as a prospective cohort study. Primary objectives were to determine anamnestic and cardiovascular factors influencing preoperative Sco2and to explore the relation between preoperative Sco2and clinical outcome (30-day and 1-yr mortality and morbidity).
After approval by the local ethical committee (Ethikkommission der Universität zu Lübeck, Lübeck, Germany), all patients scheduled for cardiac surgery with cardiopulmonary bypass (CPB) at the University of Lübeck from January 1, 2008, to December 31, 2008, were screened for participation in this prospective, observational trial. Exclusion criteria were less than 18 yr of age and planned off-pump surgery. Written informed consent was obtained from all elective and urgent patients as well as emergency patients capable of communication. In case of sedated or intubated patients scheduled for emergency surgery, consent was obtained from the next of kin and reconfirmed after recovery.
Justification of Sample Size
To our knowledge, the association between preoperative Sco2, cardiovascular and surgical risk factors, and clinical outcome has not been formally analyzed previously in larger studies. The current study may thus be regarded as a pilot trial. The sample size was adjusted to comparable studies analyzing the association between preoperative variables and outcome.
During the study period, 1,230 patients fulfilled the inclusion criteria. Two patients refused to participate in the study, 48 patients' surgery was cancelled, and no Sco2measurements were performed in 2 patients, leaving 1,178 that could be analyzed. As a result of missing either room or oxygen-supplemented Sco2concentrations or hormone values, 194 datasets were partially incomplete. Complete datasets, including hormones, were obtained from 984 patients.
Determination of Sco2
Sco2was determined by near-infrared spectroscopy with an INVOS® 4100 or 5100 cerebral oximeter (Somanetics, Troy, MI). With the exception of emergency patients and intubated patients, who were studied upon arrival in the operating room, all patients were examined preoperatively on the cardiac surgery ward. In this case, the patients were fully awake and had received neither sedating drugs nor opioids. Emergency and intubated patients received sedating or analgesic drugs, as clinically appropriate.
On the ward, Sco2was first determined in the resting state in the semirecumbent position, while the patients breathed room air, and the lowest value from the left or right hemisphere was used for determination of (Sco2min-room). Transcutaneous arterial oxygen saturation (Sao2), mean arterial blood pressure, and heart rate were determined concomitantly. Thereafter, Sco2measurements were repeated when the patients were breathing oxygen-enriched air until no further increase in Sao2was observed. The lowest Sco2value during oxygen supplementation recorded in a patient, either from the left or the right hemisphere, was determined as Sco2min-ox. In 69 patients, no Sco2measurements with oxygen were available; in these patients, Sco2min-roomvalues were used for further analysis.
Standard Risk Factors and Clinical Outcomes
Besides demographic variables serving as potential risk factors, the following specific cardiovascular risk factors were assessed:
Left ventricular ejection fraction (LVEF)
New York Heart Association grade
Estimated glomerular filtration rate
LVEF was derived from the preoperative diagnostics (levocardiography or echocardiography) and graded as: severely reduced, LVEF less than 30%; moderately reduced, 30–49%; and normal, >50%. The additive EuroSCORE was calculated according to Roques et al. 8and analyzed as a continuous variable. Estimated glomerular filtration rate was calculated from preoperative plasma creatinine by the abbreviated Modifications of Diet in Renal Disease equation.9
Clinical outcomes (30-day mortality, major complications, and duration of treatment in the high-dependency unit) were derived from the prospectively sampled cardiac surgery database. All-cause 1-yr mortality was determined from the hospital database, by contacting the patient's primary physician, the patient, or the patient's next of kin. The follow-up rate was 98.3%.
Determination of High-sensitivity Troponin T and N-terminal Pro-B-type Natriuretic Peptide
Arterial blood samples for determination of N-terminal pro-B-type natriuretic peptide (NTproBNP) and high-sensitive troponin T (hsTNT) were obtained immediately before induction of anesthesia. EDTA plasma was separated and stored at −80° for further analysis.
hsTNT was determined by the electrochemiluminescence method as described recently (ECLIA; Elecsys 2010 analyzer, Roche Diagnostics, Grenzach-Wyhlen, Germany).10The lower detection limit of this assay was 3 pg/ml. The interassay coefficient of variation was 5.4% at 28 pg/ml and 7.1% at 2,350 pg/ml.
NTproBNP was determined by an electrochemiluminescence immunoassay (Elecsys proBNP sandwich immunoassay; Roche Diagnostics) on Elecsys 2010.11The interassay coefficient of variation was 9.3% at 130 pg/ml and 14.4% at 3,890 pg/ml. The lower detection limit was 5 pg/ml.
General anesthesia was induced with etomidate and sufentanil and maintained with sevoflurane and remifentanyl, before and after CPB, and with propofol and remifentanyl, during CPB. If necessary, midazolam was added to achieve the desired anesthesia depth. Beyond standard anesthesia monitoring, including electrocardiogram, invasive arterial blood pressure, and central venous pressure monitoring, all patients were intraoperatively routinely equipped bihemispherically with Sco2sensors (INVOS® Cerebral Oximeter 5100).
Before CPB, all patients received 400 IU/kg heparin. Surgery was performed in moderate hypothermia using antegrade blood cardioplegia. Blood flow during CPB was adjusted to achieve a mean arterial blood pressure between 50–70 mmHg, a mixed venous oxygen saturation—measured at the inflow of the CBP circuit—higher than 70%, and relative Sco2concentrations higher than 80% of the preoperative baseline determined in the operating room when breathing room air. To achieve this goal, norepinephrine and nitroglycerin were applied as a bolus or a continuous infusion. Hematocrit was adjusted between 26% and 29%. Acid-base balance was performed after α-stat blood gas principles. After weaning from CPB, protamine was applied accordingly.
Of the 1,178 patients, 800 (67.9%) underwent CABG surgery (isolated CABG, 571; CABG plus aortic valve surgery, 151; CABG plus other valve surgery, 62; CABG plus ascending aortic surgery, 4). Of the total number of patients, 378 (32.1%) underwent surgery other than CABG (aortic valve surgery, 174; mitral valve surgery, 37; tricuspid valve surgery, 7; combined valve surgery, 127; major thoracic vascular surgery, 60; other on-pump procedures, 27).
Analyses were performed in the total cohort and in a group of 102 high-risk patients with an additive EuroSCORE more than 10.12No adjustments for multiple testing were made.
Data were analyzed by MedCalc 11.3.3 statistical software package (MedCalc Software bvba, Mariakerke, Belgium). Because most of the relevant data were not normally distributed after Kolmogornov–Smirnov testing, all data, if not stated otherwise, are given as median and 25% of 75% quartiles. Statistical significance was assessed at the 5% level (P < 0.05 is statistically significant).
To allow better comparisons with the additive EuroSCORE, analyses regarding mortality were primarily calculated for 30-day mortality. One-yr mortality was used for construction of the Kaplan–Meier survival curves and logistic regression analyses.
Postoperative morbidity was defined as a combined endpoint of (A) more than 1 point in the major adverse events and complications score (MaCS) according to Schön et al. 7(need of renal replacement therapy, reintubation, stroke [Rankin disability score13>1], and low-cardiac output syndrome) or (B) need for high-dependence unit (intensive care unit plus intermediate care) treatment of at least 10 days.
Comparisons between groups for univariate predictors of outcome were performed by a two-sided chi-square test with Yates correction for categorical variables and Mann–Whitney test or Kruskal–Wallis test, as appropriate, for continuous variables.
Correlation analyses were performed by the Spearman rank correlation test for continuous variables or Kendall's τ for scaled variables, as appropriate. Cutoff values for mortality and morbidity were derived from receiver-operating curve (ROC) analyses. The lowest cutoff values for Sco2min-oxderived from the ROC analysis was used for construction of a Kaplan–Meier survival curve with log-rank test statistics.
Backward logistic regression analyses were performed by entering additive EuroSCORE, Sco2min-ox, and CPB duration in the model. For the latter variables, the cutoff values derived from the ROC analyses were used. EuroSCORE was added as a graded variable with the different risk categories (0–2; 3–5, 6–10, and >10) graded as 1–4.
Analyses were performed in the total cohort of patients with and without entering the cutoff concentrations of hsTNT and NTproBNP. With respect to the low event rate for 30-day mortality, additional analyses were performed using 1-yr mortality as the dependent variable.
Demographics and Surgical Course
Of the total number of patients, 70.5% were men. The median age was 68 yr (59–74 yr), weight was 81 kg (72–91 kg), height was 172 cm (166–178 cm), and body mass index was 27 (24–32). The median American Society of Anesthesiology classification was 3 (3–3), and median additive EuroSCORE was 5 (3–8). Of the total number of patients, 484 (41%) were graded as New York Heart Association grade III/IV, and 694 (59%) were in New York Heart Association grade I/II.
Cardiovascular risk factors, preoperative medication, baseline hemodynamics, hematocrit, estimated glomerular filtration rate, hsTNT, and NTproBNP are presented in table 1.
A recorded 1,058 patients (90%) were in elective surgical status, with 120 (10%) of these cases classified as urgent or emergency surgery. A recorded 635 patients (54%) suffered from angina pectoris, 215 (18%) had a main left stem coronary stenosis, 295 (25%) had a history of myocardial infarction, 108 (9%) were scheduled for redo surgery, and 33 patients (2%) had an active endocarditis. Before surgery, 61 (5%) patients needed circulatory support (i.e. , inotropic and vasopressor support or intraaortic counterpulsation). Main variables of the surgical course and outcomes are presented in table 1. Thirty-day and 1-yr mortality and major morbidity (at least two major complications and/or a high-dependency unit stay of at least 10 days) were 3.5%, 7.7%, and 13.3%, respectively.
The subgroup of high-risk patients with a EuroSCORE greater than 10 consisted of 102 patients with a median additive EuroSCORE of 12 (11–14). The patients in this group were older (74 yr [67–78 yr]), weighed less (75 kg [67–97 kg]), and were smaller (170 cm [160–178 cm]; all: P < 0.05) compared with the total cohort.
The surgical mix was comparable with the total group. In this group, 60.8% of patients underwent CABG surgery and 39.2% were categorized as NO-CABG. Of these high-risk patients, 57.8% were graded as urgent or emergent. Duration of surgery was 308 min (261–410 min), CPB and aortic cross-clamp time were 150 (116–200 min) and 111 min (78–145 min), respectively. Thirty-day and 1-yr mortality were 15.7% and 27.5%. Of these patients, 40.2% fulfilled the criteria of the combined morbidity endpoint. NTproBNP concentrations were 2,739 pg/ml (1,074–29,346 pg/ml) and hsTNT concentrations were 72.5 ng/ml (17.2–376 ng/ml). With the exception of the surgical mix, all depicted variables were significantly (P < 0.05) different from the total cohort.
Preoperative Cerebral Oxygen and Arterial Saturation
Sco2values available with room air were 1,123, and Sco2values when breathing oxygen-enriched air were 1,109. The Sco2min-roomin the total cohort when patients were breathing room air with a median Sao2of 97% (95–98%) was left 62% (57–67%) and right 62% (56–67%). The Sao2when breathing oxygen-enriched air was 100% (100–100%), and Sco2with oxygen was left 66% (61–71%) and right 66% (61–71%).
Sco2min-oxderived from the measurements during oxygen supplementation in 1,109 patients and from room air measurements in 69 patients was 64 (55/69). A histogram of Sco2min-oxvalues in the total cohort is given in figure 1.
In high-risk patients with an additive EuroSCORE more than 10, Sao2with room air was 96% (94–97%). Mean arterial blood pressure and heart rate were 80 mmHg (80–100 mmHg) and 75 beats/min (66–91 beats/min), respectively. Sco2min-roomwas 54% (49–60%), and Sco2min-oxwas 60% (52–65%). All oximetry results in the high-risk group were significantly different from the total cohort.
Factors Influencing Sco2
Table 2shows that Sco2min-oxwas correlated with various demographic and physiologic variables, especially age, gender, body mass index, American Society of Anesthesiology grade, additive EuroSCORE, LVEF, glomerular filtration rate, and hematocrit.
Analysis of Sco2min-oxin different EuroSCORE groups showed significant differences between the groups with higher Sco2concentrations in the decreased additive EuroSCORE groups (fig. 2). Comparably, patients with a decreased LVEF, a higher NTproBNP, and a higher hsTNT had lower Sco2min-oxthan the healthier groups of patients (fig. 2).
Sco2and Clinical Outcomes
Univariate analyses of patients with postoperative mortality and morbidity revealed that patients with an adverse outcome had decreased preoperative Sco2levels, a higher additive EuroSCORE, a decreased preoperative hematocrit, and a longer duration of surgery and CPB (table 1).
ROC analysis of Sco2min-oxand 30-day mortality revealed an area under the curve (AUC) of 0.71 (95% CI, 0.68–0.73; P < 0.0001) and a cutoff value of ≤51% (sensitivity 41.5%, specificity 93.6%) in the total cohort and an AUC of 0.77 (95% CI, 0.69–0.86; P < 0.0001) and a cutoff value of ≤53% (sensitivity 75.0%, specificity 79.1%) in patients with a EuroSCORE more than 10. The respective results for the combined morbidity endpoint (a MaCS of at least two and/or a high-dependency unit time of at least 10 days) revealed an AUC of 0.68 (95% CI, 0.65–0.70; P < 0.0001) in the total cohort. The cutoff value was a Sco2min-oxof ≤60% with a sensitivity of 56.1% and a specificity of 71.4%.
To rule out a possible confounding effect of intubation and controlled ventilation in emergency patients, the ROC analyses were also performed in the subgroup of elective patients. The analysis of Sco2min-oxand 30-day mortality revealed an AUC of 0.66 (95% CI, 0.63–0.69) with a cutoff value of ≤53%, a sensitivity of 40.9%, and a specificity of 90.7% (P = 0.018). The comparable analysis in the patients with emergency status revealed an AUC of 0.74 (95% CI, 0.68–0.8) with a cutoff Sco2min-oxof ≤50%, a sensitivity of 47.4%, and a specificity of 91.5% (P = 0.0002).
To determine whether the capability of increasing Sco2concentrations during oxygen supplementation in patients with a low Sco2min-roomhad prognostic relevance, patients presenting with a Sco2min-roomof ≤51% and showing an increase to Sco2min-oxconcentrations higher than 51% during oxygen application were compared with those patients that remained below 51% Sco2min-ox. Of 142 patients, 81 with a Sco2min-roomof ≤51% responded to oxygen by increasing ScO2min-oxto concentrations higher than 51%; in 61 patients, Sco2min-oxremained lower or equal than 51%. Mortality and morbidity were significantly higher in the nonresponders than in the responders (30-day mortality: 16.4% vs. 2.5%, P = 0.0081, combined morbidity endpoint; a MaCS of at least two and/or a high-dependency unit time of at least 10 days: 37.7% vs. 14.8%, P = 0.0033).
Predictive Accuracy of Sco2
versus Additive EuroSCORE
Comparative ROC analyses of additive EuroSCORE with respect to 30-day mortality and the combined morbidity endpoint in the total cohort revealed an AUC of 0.82 (95% CI, 0.8–0.84%) for mortality and an AUC of 0.77 (95% CI: 0.74–0.79%) for morbidity. A comparison with Sco2minshowed that the EuroSCORE had a better accuracy in predicting 30-day mortality (difference between areas, 0.11; P = 0.015; fig. 3) and morbidity (difference between areas, 0.09; P < 0.001) in the total cohort. Comparable results were obtained in patients with an additive EuroSCORE below 10 (data not shown).
In the high-risk group with an additive EuroSCORE more than 10, Sco2min-oxhad a better accuracy for 30-day mortality than the EuroSCORE (EuroSCORE, 0.54 [95% CI, 0.43–0.64%]; Sco2min-ox, 0.79 [CI 95%, 0.7–0.86%]; δ AUC 0.25, P = 0.0044) (fig. 3). The predictive capacity for morbidity in the high-risk group was not different between the EuroSCORE and the Sco2min-ox(EuroSCORE, 0.61; [95% CI, 0.51–0.71%]; Sco2min-ox: 0.64 [CI 95%, 0.57–0.76%]; δ AUC 0.06, P = n.s.).
Predictive Accuracy of hsTNT and NTproBNP
ROC analyses of hsTNT and 30-day mortality revealed an AUC of 0.8 (95% CI, 0.77–0.82%; P < 0.001) and an optimal cutoff value of 28 pg/ml with a sensitivity of 81.8% and a specificity of 73.8% in the total cohort. The respective results in the high-risk group (EuroSCORE more than 10) were: AUC of 0.71 (95% CI: 0.59–0.81%), P = 0.002, a cutoff of 37.1 pg/ml, a sensitivity of 100%, and a specificity of 50%.
ROC analysis of the predictive capacity of hsTNT for the combined morbidity endpoint showed an AUC of 0.72 (95% CI, 0.69–0.75%), P < 0.001, and an optimal cutoff of 21.8 pg/ml with a sensitivity of 62.7%, and a specificity of 71.5%. The respective analysis in the high-risk group failed to reach statistical significance.
ROC analysis of NTproBNP and 30-day mortality showed an AUC of 0.78 (95% CI, 0.76–0.81%), P < 0.001, and a cutoff concentration of 1,743 pg/ml with a sensitivity of 75.8% and a specificity of 79.9% in the total cohort. The ROC analysis of NTproBNP and mortality in the high-risk group was not significant. ROC analysis of NTproBNP and morbidity revealed an AUC of 0.7 (95% CI, 0.67–0.73%), P < 0.001, and a cutoff concentration of 1,061 pg/ml with a sensitivity of 61.2% and a specificity of 74.4%.
Comparative analyses on the accuracy of NTproBNP, hsTNT, and Sco2min-oxwere performed in the subgroup of patients with complete hormone analysis and revealed no significant differences in the accuracy for predicting 30-day and 1-yr mortality in the total cohort (data not shown).
Backward logistic regression analyses for 30-day and 1-yr mortality was performed using a model with the EuroSCORE graded in different risk categories (0–2, 3–5, 6–10, >10) and the lowest cutoff value derived from the different ROC analyses for Sco2min-ox(≤50%). As a marker for the severity of the surgical insult, CPB time was added to the model. The cutoff value for 30-day mortality derived from ROC analysis for CPB time was more than 114 min with an AUC of 0.76, a sensitivity of 82.9%, and a specificity of 57.2% (P < 0.0001). To allow easier comparisons in future studies, the CPB time cutoff was set to more than 120 min. The 30-day mortality rates in the different EuroSCORE categories were: EuroSCORE 0–2, 0.3%; EuroSCORE 3–5, 1.3%; EuroSCORE 6–10, 5.2%; EuroSCORE more than 10, 16.7%.
With respect to the fact that preoperative determination of NTproBNP and hsTNT cannot be considered routine in many European heart centers, calculations were either performed with or without inclusion of NTproBNP and hsTNT. The results of the logistic regression analyses are given in table 3, showing that a Sco2min-oxequal or less than 50% is an independent predictor of 30-day mortality, if conventional risk factors are used, and of 30-day, as well as 1-yr, mortality, if NTproBNP and hsTNT are entered into the model. The category EuroSCORE 3–5 was not included in the regression models because of the low event rate (5 deaths per 401 patients).
Kaplan–Meier Survival Analyses
The effects of Sco2on long-term mortality using a Sco2min-ox≤50% as a cutoff value was analyzed by Kaplan–Meier statistics. Log-rank test revealed a significant effect in the total cohort as well as in the high-risk group with a EuroSCORE more than 10 (fig. 4).
Increasing evidence suggests that decreased Sco2during surgery is associated with unwarranted neurologic effects and a complicated postoperative course.1,3Most studies in this field have only focused on the relative changes of Sco2during cardiac6,14,15and noncardiac surgery.16,17To what extent decreased preoperative Sco2concentrations are related to postoperative complications and mortality has not been formally analyzed in a large cohort of patients.
The findings of the current study, showing that preoperative Sco2levels are closely related to relevant measures of cardiopulmonary function, postoperative morbidity, and short- and long-term mortality, may help to better understand the role of perioperative changes in Sco2and may have implications for improving risk stratification in cardiac surgery.
From a technical point of view, Sco2, determined by near-infrared spectroscopy, is reflective of tissue oxygen saturation in a small sample of the frontal cortex, and, with respect to the distribution of arterial and venous blood, weighted by the venous compartment.1,18Various experimental and clinical studies have clearly shown that Sco2determined by near-infrared spectroscopy is directly related to cerebral blood flow and jugular venous oxygen saturation (Svo2).19,20
Current evidence supports that Sco2also indirectly reflects the adequacy of the systemic circulation and cardiopulmonary function. In line with other studies comparing Sco2and mixed Svo2in children,21we were able to show a close correlation between Sco2and Svo2in extubated patients after cardiac surgery.**Comparably, despite the absolute levels of Sco2and Svo2that were only moderately correlated, Dullenkopf et al. showed that trends in Sco2reflect changes in Sco2in intubated patients immediately after cardiac surgery.22Skhirtladze et al. observed that patients with severely reduced LVEF being scheduled for implantation of a defibrillator cardioverter system had lower Sco2concentrations than patients with a preserved LVEF.23Madsen et al. showed that patients presenting with acute heart failure had lower Sco2concentrations than healthy control patients and that Sco2increased during heart failure therapy.24Koike et al. showed that the decrease in Sco2during exercise in patients with coronary artery disease had prognostic effect and was associated with an unfavorable outcome in these patients.25Paquet et al. observed that baseline Sco2was related to LVEF in patients undergoing cardiac surgery and that Sco2had an acceptable accuracy to identify left ventricular dysfunction in these patients.26
The findings of the current study, showing significant correlations between NTproBNP, hsTNT, LVEF, and Sco2in a large patient sample, support the concept that Sco2is influenced by myocardial function and the performance of the cardiopulmonary system.
This may have clinical and scientific implications. First, most guidelines on Sco2monitoring for guiding hemodynamics and CPB management during cardiac surgery suggest the adjustment of therapy according to relative changes from the preoperative baseline (i.e. , to maintain relative Sco2concentrations higher than 80% of baseline).27Based on our findings, this concept may be appropriate for patients starting with a Sco2in a healthy range but may be inadequate for patients presenting with decreased preoperative Sco2as an expression of severe cardiopulmonary dysfunction. Second, these findings offer scientific perspectives for using Sco2as a noninvasive technology for guiding therapy in patients with heart failure as already suggested by Madsen et al. more than 10 yr ago.24
The accuracy of Sco2for predicting 30-day mortality in the total cohort was lower than the additive EuroSCORE but comparable with established (NTproBNP)28,29and emerging (hsTNT)10,30humoral markers of postoperative risk in cardiac surgery. As derived from the ROC and the Kaplan–Meier analyses, the accuracy of Sco2for predicting short- and long-term prognosis in high-risk patients is clearly superior to the EuroSCORE. It is noteworthy that patients with a EuroSCORE more than 10 and a Sco2min-oxequal or less than 50% had a 1-yr mortality that was twice as high as in high-risk patients with preserved—more than 50%— Sco2concentrations. This is a remarkable finding because risk stratification in high-risk cardiac surgery patients by EuroSCORE, as frequently performed in most European heart centers, is limited by the fact that the additive and the logistic EuroSCORE overestimate mortality in this group of patients.31,32Consequently, our findings, if reproduced prospectively, may have relevant implications for improving risk stratification in high-risk cardiac surgery patients.
To rule out a possible confounding effect of arterial hypoxia on the results of Sco2monitoring, we measured Sco2with room air and during oxygen insufflation. The difference between the median results with both measures was 4% and comparable with the 3% difference in transcutaneous oxygen saturation achieved with application. Numerically, the accuracy for predicting 30-day mortality was not different if the Sco2values derived with room air or during oxygen supplementation were used (data not shown). However, an analysis of whether the capability to increase Sco2during oxygen supplementation in patients with a low Sco2during room air had prognostic relevance revealed that nonresponders (i.e. , patients that failed to increase Sco2beyond the cutoff value) had a significantly higher mortality and morbidity than those showing an increase beyond the Sco2cutoff. This suggests that if Sco2will be used for risk stratification, this parameter should ideally be determined during oxygen supplementation.
Besides the EuroSCORE, various other risk stratification models are used in cardiac surgery.33Currently, most of these models, comparable with the EuroSCORE, have been developed and validated in patients undergoing CABG surgery, and the predictive accuracy for patients undergoing other types of surgery is often limited. In contrast, the Society of Thoracic Surgeons score is available in different versions for various kinds of cardiac procedures. This score has an excellent predictive capacity but needs a large number of variables to be entered.34–36Consequently, at least in Europe, it is less frequently used than the EuroSCORE.
Focusing more on physiologic than historical variables, Ranucci et al. have shown that a simple score consisting of the variables age, creatinine, and left ventricular ejection fraction has an accuracy for 30-day mortality that is comparable with the additive and the logistic EuroSCORE.37This suggests that risk stratification based on age and physiologic variables is feasible and may at least be as predictive as a classic risk score.
Our findings, regarding the discriminatory capacity of NTproBNP for predicting 30-day mortality, are in line with previous observations in cardiac28,29as well as noncardiac patients.29The accuracy of NTproBNP was comparable with the additive EuroSCORE and Sco2min. The latter also holds true for hsTNT, an observation that is remarkable in so far as this preclinical test has not been used in patients undergoing cardiac surgery before. It is also of note that the hsTNT cutoff for 30-day mortality was highly comparable with the one derived from a large trial in medical patients with heart failure,10further suggesting that hsTNT serves not only as a marker of myocardial ischemia but, comparable with NTproBNP, as a measure of global cardiopulmonary dysfunction.
The current study has limitations.
The short-time mortality in the study population was low. Thus, the results presented, despite being statistically significant, are prone to potential error and require prospective confirmation. However, our findings are substantiated by the fact that when analyzing 1-yr mortality, the number of patients with an adverse outcome was more than twice as high and remained statistically significant not only when focusing on the total cohort but also on the subsets of patients with different surgical priority.
The Sco2monitors used in the current study (INVOS® 4100 and 5100) only have a Federal Drug Administration approval for trend monitoring of Sco2and not for absolute values. Recently, another Sco2monitoring system (Fore-Sight®; CAS Medical Systems, Branford, CT) has been introduced to the market that, by using four instead of three wavelengths for determination of oxygenated and deoxygenated hemoglobin and laser light instead of a light emitting diode, has been suggested to be capable of measuring absolute Sco2 concentrations.3No comparative data of Sco2concentrations derived by these different monitors are available, thus it remains currently speculative whether one system is more precise than the other. However, it is likely that any measure improving the reliability of the near-infrared spectroscopy technology will also improve the accuracy of Sco2for risk stratification.
Unfortunately, Sco2concentrations with oxygen were only available for 1,109 patients and were missing in some patients that have died. However, with respect to the observation that patients with a poor prognosis showed only minor increases of Sco2during oxygen supplementation, we have used the Sco2min-roomvalues from the missing 69 patients for calculation of Sco2min-oxto allow calculations on the maximal number of patients with adverse events.
The modeling strategy of the current study was designed to explore associations, and many of the analyses are based on post hoc and subgroup analyses of the primary dataset. Consequently, the results of the current study require prospective replication in an independent sample.
In conclusion, the current study shows that preoperative Sco2levels determined by near-infrared spectroscopy are related to objective measures of cardiopulmonary function and that low preoperative Sco2concentrations are associated with an adverse perioperative course. A Sco2min-oxequal or less than 50% is an independent predictor of short- and long-term mortality in patients undergoing on-pump cardiac surgery and might serve as a refined marker for preoperative risk stratification in cardiac surgery patients.
The members of the cardiac anesthesia unit and the personnel of the Department of Clinical Chemistry (University of Lübeck, Lübeck, Germany) are thanked for their help in data and blood sample acquisition.