Central Venous and Mixed Venous Oxygen Saturation in Critically Ill PatientsLadakis C.a · Myrianthefs P.a · Karabinis A.b · Karatzas G.b · Dosios T.b · Fildissis G.a · Gogas J.b · Baltopoulos G.a
aAthens University School of Nursing Intensive Care Unit at Agioi Anargyroi Cancer Hospital of Kifissia, and bAthens University School of Medicine A’ Surgical Department at Laikon Hospital of Athens, Athens, Greece Corresponding Author
Background: Although mixed venous O2 saturation (SvO2) accurately indicates the balance of O2 supply/demand and provides an index of tissue oxygenation, the use of a pulmonary artery (PA) catheter is associated with significant costs, risks and complications. Central venous O2 saturation (ScvO2), obtained in a less risky and costly manner, can be an attractive alternative to SvO2. Objectives: To investigate whether the values of ScvO2 and SvO2 are well correlated and interchangeable in the evaluation of critically ill ICU patients and to create an equation that could estimate SvO2 from ScvO2. Methods: Sixty-one mechanically ventilated patients were catheterized upon admission and ScvO2 and SvO2 values were simultaneously measured in the lower part of the superior vena cava and PA respectively. Results: SvO2 was 68.6 ± 1.2% (mean ± SEM) and ScvO2 was 69.4 ± 1.1%. The difference is statistically significant (p < 0.03). The correlation coefficient r is 0.945 for the total population, 0.937 and 0.950 in surgical and medical patients, respectively. In 90.2% of patients the difference was <5%. When regression analysis was performed, among 11 models tested, power model [SvO2 = b0(ScvO2)b1] best described the relationship between the two parameters (R2 = 0.917). Conclusions: ScvO2 and SvO2 are closely related and are interchangeable for the initial evaluation of critically ill patients even if cardiac indices are different. SvO2 can be estimated with great accuracy by ScvO2 in 92% of the patients using a power model.
Copyright © 2001 S. Karger AG, Basel
Mixed venous oxygen saturation (SvO2) obtained by an oximetric pulmonary artery (PA) catheter indicates the balance between oxygen supply and demand and thereby provides an index of tissue oxygenation [1, 2, 3]. However, the use of a PA catheter is associated with increased risks, severe complications and increased health care costs . The risk/benefit ratio is under vigorous debate [8, 9, 10]. PA catheter use is associated with a complication rate of up to 33% and a mortality rate of up to 1.5% which further contribute to the increased equipment and personnel costs of the technique [5, 6, 11]. On the other hand, central venous catheters (superior vena cava; SVC) are much less costly and have a much lower risk of significant complications of up to 4.4% . Also, central venous catheters are routinely used in ICU patients for central venous pressure monitoring, fluid resuscitation and the administration of vasoactive or other drugs and total parenteral nutrition. Measurement of central venous oxygen saturation (ScvO2) using central venous catheters seems to be a simple method and an attractive alternative to measurement of SvO2 for the initial evaluation and monitoring of critically ill patients, since it can be obtained in an easier, less risky and less costly manner.
The aim of this prospective study was to ascertain whether ScvO2 correlates sufficiently well with SvO2, and can therefore replace the measurement of SvO2 in the initial evaluation of critically ill patients on admission to the ICU, and to find an equation for the accurate calculation of SvO2 from ScvO2.
material and methods
In this prospective study, 61 patients of either sex (43 males) requiring measurement of PA pressure and cardiac output (CO) were catheterized with an Opticath, (PA Catheter P 7110, 7.5 F, Abbot Laboratories). The Human Investigation Committee of our Institution and the University of Athens, Faculty of Medicine, approved the study protocol. The aim was to measure the values of SvO2 and ScvO2 within a short interval of time in critically ill patients upon their admission to the ICU. During the measurements, all patients were receiving Assist Control (A/C) mechanical ventilation, were sedated with midazolam (0.1 mg/kg/h) and paralyzed with atracurium besylate (0.7 mg/kg/h). Before the insertion, the PA catheter was calibrated in vitro according to the manufacturer’s instructions. The PA catheter was inserted following standard sterile procedures and universal safety precautions and then advanced at first in the lower part of the SVC. Correct positioning in the SVC was confirmed by drawing back the catheter by 2 cm after obtaining a right atrial pressure waveform. Immediately thereafter, the catheter was advanced in the PA with the balloon inflated with 1.5 ml of air until a typical wedge pressure waveform was obtained. A chest X-ray confirmed the correct position of the PA catheter tip. The catheter was connected to a venous oxygen saturation monitor (Oximetric 3, Abbot Laboratories) which has the capacity to immediately and continuously display venous oxygen saturation (either central-SVC or mixed venous-PA).
The patients’ characteristics are presented in table 1. The site of the operation was the abdomen (n = 22), the thoracic cavity (n = 4), the skull for cerebral hemorrhage (n = 4) and lower extremity large vessels (n = 1). Medical patients presented with respiratory failure due to COPD exacerbation (n = 19), cardiogenic pulmonary edema (n = 5), ARDS due to intra-abdominal sepsis (n = 4), laryngeal edema (n = 1) and near drowning (n = 1). Illness severity scoring was SAPS II = 46 ± 3.2 and cardiac index (CI) value of 3.94 ± 0.18 liters/ min/m2.
Table 1. Characteristics, severity of illness and hemodynamic status on admission
With each of the 61 patients in a stable condition and without any other medical (drug) intervention we noted in succession, at first the ScvO2 value in the lower part of the SVC and immediately thereafter, by advancing the catheter in the PA within a very short interval of time, the SvO2 value on the screen of the monitor. Real time interval needed in practice in the 61 patients ranged from 10 to 30 s. The values of the two parameters as well as CO and CI were recorded for each of the 61 patients.
The values of the two parameters, age, severity of illness score and CI are expressed in mean value ± SEM. Statistical analysis was performed by paired sample t test, linear regression and Pearson correlations. All statistics were performed by using the SPSS (version 8.0) statistical package.
In order to find a mathematical model that could be used to estimate SvO2 from ScvO2 minimal error, a regression analysis was also performed. The model (among 11 tested: linear, logarithmic, inverse, quadratic, cubic, power, compound, S, logistic, growth, exponential) found to best interpret the relation between the values of the two parameters in terms of R2 was that of a ‘power model’, which has the following expression: SvO2 = b0(ScvO2)b1, where the values of the coefficients were calculated to be b0 = 1.1612 and b1 = 0.9617 for the total population. This model equation has a value of R2 = 0.917, that is 92% of the variance of SvO2 values can be explained by the ScvO2 values. Similarly, the power model can be used for subpopulations in order to find the clinical significance of differences between the value of SvO2 and ScvO2. The values of b0 and b1 and the values of R2 are shown in table 2.
Table 2. The values of b0 and b1 and the values of R2 [SvO2 = b0(ScvO2)b1]
We obtained data concerning the hemodynamic status (e.g. CI, table 1) and the desired variables (SvO2 and ScvO2, table 3) in 61 patients within the 1st hour after admission to the ICU. Descriptive statistics and correlations are presented in table 3. There was a statistically significant difference between mean values for the total population (p < 0.03) and surgical patients (p < 0.035) but not for medical patients (p < 0.306). The correlation between SvO2 and ScvO2 for the total patient population is shown in figure 1.
Fig. 1. Correlation between ScvO2 and SvO2. A significant correlation was found.
Table 3. Descriptive statistics and correlation coefficients upon admission to the ICU
The difference between the two parameters for the total population was 0.8% (range 0.5–1.3 among subpopulations). Table 4 lists the relative frequencies of the differences between SvO2 and ScvO2 for the total population. In 90.1% of the patients, the difference was less than 5 units of saturation (%).
Table 4. Frequency of differences between SvO2 and ScvO2 expressed in magnitude of difference
The correlation coefficient (table 3) between SvO2 and ScvO2 in the total population was 0.945 (range 0.846–0.959 in subpopulations).
The influence of CI divided into three categories (low, medium and high) on the correlation between the two variables was also examined and is summarized in table 5. The correlations ranged between 0.846 in high CI and 0.974 in low CI values. All values of r are significant at the level of 0.01.
Table 5. Correlation between SvO2 and ScvO2 depending on CI
Figure 2 shows the line connecting the observed values (dotted line) and the line (continuous) obtained using the power model equation SvO2 = b0(ScvO2)b1 in the total population.
Fig. 2. Regression analysis with the power model in the total population. The dotted line represents observed values and the continuous line represents calculated values using the power model.
According to our results there is a high correlation between the two parameters (r 0.945) on admission to the ICU (table 3, fig. 1). That is, the strength of the linear relationship between the two variables is positive. This high value of the correlation coefficient allows us to suggest that ScvO2 parallels SvO2 sufficiently and thus the two parameters are interchangeable in clinical practice. That is, we can use ScvO2 as a mirror of SvO2 for the initial evaluation of critically ill patients upon admission to the ICU. In 90.2% of the patients, the values of the two variables differed by less than 5% (table 4), which is an acceptable difference that neither influences patient evaluation nor therapeutic choices. Also, regression analysis allows us to exactly explain the 92% of the variance of SvO2 from ScvO2 using the power model (R2 = 0.917). Finally, the influence of CI on the correlation between the two variables is minimal (table 5), especially in the case of low and medium CI values (r = 0.974 and 0.968, respectively) but also in patients with high CI (r = 0.846).
The correlation between SvO2 and ScvO2 and the usefulness of ScvO2 measurements has been evaluated in the past with different conclusions in different clinical settings (table 6). Close correlation was found over a wide range of CO in patients with myocardial infarction admitted in a coronary care unit, and ScvO2 measurement was helpful in assessing the direction of cardiac function and CO changes and monitoring drug therapy [12, 13]. Also, in an experimental study in dogs, Reinhart et al.  found a high correlation between the two parameters during changes in oxygen supply/demand (r = 0.91–0.97) and that changes in ScvO2 closely paralleled SvO2 changes. The same authors suggested that ScvO2 warrants further consideration for patient evaluation and monitoring of trends in O2 supply and demand. Berridge  found a high correlation (r = 0.93) between the two parameters even when the exact position of the central venous catheter was unknown and that the effect of CO on the correlation is minimal. He author suggested that ScvO2 is very useful in the evaluation of patients and the estimation of SvO2, oxygen extraction ratio and arteriovenous oxygen content difference. These findings concerning the influence on CI on the correlation between the two variables are compatible with our findings.
Table 6. Brief survey of the literature on the correlation of SvO2 versus ScvO2
However, others found that there is a poor correlation between the two parameters and questioned the usefulness of ScvO2 measurement, particularly in patients with shock and heart failure [15, 16]. Scheinman et al.  found that ScvO2 accurately reflects (r = 0.99) SvO2 in seriously ill patients only when the patient is not in shock or heart failure. However, they found that changes in ScvO2 are more important than absolute values and that there is a good correlation between changes in SvO2 and respective changes in ScvO2. Finally, they suggested that central venous catheters should be positioned in the superior portion of the right atrium. These results are supported by Lee et al. , who found a good correlation in normal subjects (r = 0.88) but not in patients in shock (r = 0.73). Martin et al.  studied 7 critically ill patients with 448 consecutive measurements of the two parameters and found that there is a poor correlation between them during periods with or without medical intervention (r = 0.48 and 0.62, respectively) and they concluded that the two parameters are not interchangeable. They also found that the difference between the two parameters was >5% in 50% of the measurement, that abrupt changes in SvO2 were not detected by ScvO2 in 18% of the measurements and that there is poor correlation between changes in the two parameters during periods without (r = 0.70) and with therapeutic interventions (r = 0.77). A relatively poor correlation (r = 0.75) was reported by Faber  in a study conducted in a heterogeneous group of 24 critically ill patients . However, he found a substantial improvement in correlation (r = 0.94) when he compared oxygen content to oxygen saturation (by multiplication of saturation by a common hemoglobin value) and he concluded that the results are mathematically coupled to a degree that precludes final conclusions and thus necessitates further prospective evaluation. Finally, Vincent , using the data of Martin et al.  concluded that ScvO2 should not be considered as a reliable substitute for SvO2. However, the very small number of patients of this study and thus the dependent nature of the measurements make these results inaccurate for final results and conclusions. A brief synopsis of the correlations and significance of the SvO2 and ScvO2 literature is shown in table 6.
It is also well established that central venous blood (superior or inferior vena cava blood) is not mixed venous blood, because there can be substantial variations in the oxygen content of the upper and lower part of the body . In healthy subjects  ScvO2 is normally lower than SvO2, because the lower extraction ratio by the kidneys leaves a high oxygen content in the inferior vena cava. However, in low flow states or in shock, this gradient can be reversed by redistribution of blood flow associated with greater reduction in renal and splanchnic blood flow [15, 19]. In the population we studied we found a reversed gradient with a value of ScvO2 0.8% larger than that of SvO2. This finding is in accordance with the literature but recorded in patients without any evidence of shock or low flow state. Although this reversed difference was statistically significant (p < 0.03), the correlation between the two variables is extremely high, suggesting that the two variables were sufficiently parallel as shown in figure 1. Also, a reversed gradient was surprisingly observed in patients with medium (by +1.1) and high (by +0.9) CI values and not in those with low CI values in whom SvO2 was 0.6% larger than ScvO2.
In spite of the discrepancies reported in the literature, our results suggest that the two parameters are closely correlated and that ScvO2 may be used as a mirror of SvO2 for the initial evaluation of critically ill patients. In other words, the two parameters are interchangeable even in the case of different CI values. Also, using the power model equation of regression analysis we can estimate SvO2 from ScvO2 with great accuracy in 92% of the patients (R2 = 0.917). Finally, the differences between ScvO2 and SvO2 are compensated by the lesser risks and costs associated with central venous catheterization in comparison with right heart catheterization [4, 5, 6, 7, 8, 9, 10, 11]. Of course, more clinical studies are needed to confirm the results of our study especially in subpopulations including patients with sepsis, head trauma, ARDS, cardiogenic pulmonary edema.
We would like to acknowledge Prof. Fotis Georgiakodis for the statistical analysis of the data.
Pavlos Myrianthefs, MD
Aegiou 34, Ampelokipoi
GR–11527 Athens (Greece)
Tel. +30 1 7791635, Fax +30 1 6280702
Received: Received: July 10, 2000
Accepted after revision: December 28, 2000
Number of Print Pages : 7
Number of Figures : 2, Number of Tables : 6, Number of References : 20
Respiration (International Review of Thoracic Diseases)
Founded 1944 as ‘Schweizerische Zeitschrift für Tuberkulose und Pneumonologie’ by E. Bachmann, M. Gilbert, F. Häberlin, W. Löffler, P. Steiner and E. Uehlinger, continued 1962–1967 as ‘Medicina Thoracalis’
Vol. 68, No. 3, Year 2001 (Cover Date: May-June 2001)
Journal Editor: C.T. Bolliger, Cape Town
ISSN: 0025–7931 (print), 1423–0356 (Online)
For additional information: http://www.karger.com/journals/res