Central Venous Oxygen Saturation above 75% on Day Three of Septic Shock is Associated with Tripled Mortality

Background: Central venous oxygen saturation (ScvO2) is commonly used to identify global oxygen uptakedelivery mismatches. A level above 70% was the declared goal in early resuscitation of septic shock for over a decade. Recent evidence leads to doubts and levels higher than 80% may represent harmful conditions. This study ́s aim was to identify favourable ScvO2 levels in treatment of septic shock. Methods: Electronic data of patients suffering from septic shock who have been admitted to the surgical intensive care unit of a university hospital were analysed surveying a period of six years with focus on the association of ScvO2 levels with in-hospital mortality. Results: Data from 238 individuals were included. No difference was found comparing initially measured values of ScvO2 of survivors to non-survivors. Patients whose levels of ScvO2 never exceeded 70% (n=28) had a higher mortality rate (73.2% vs. 54.3%, p<0.05). On day three patients with values above 75% (n=32) had higher mortality rates (59.4% vs. 38.5%, p<0.05). A mortality rate of 100% was detected if ScvO2 levels exceeded 84% (n=6). Conclusions: ScvO2 develops from a therapy guiding parameter to a prognostic marker. Reaching levels of at least 70% within the first 72 h of disease is favourable in regard to prognosis. Exceeding 75% after day two is associated with higher mortality. These findings require further confirmation. At this point it can be assumed that a varying, time-dependent ScvO2 approach might be required for clinical decision-making.


Introduction
Despite intensive efforts of improving therapy in treatment of septic shock mortality rates are still reported about 36.5% [1]. A number of 5.3 million deaths annually worldwide is estimated [2]. Treatment recommendations [3,4] focus on infection control and cardiorespiratory resuscitation to restore oxygen supply, optimize tissue perfusion, avoid hypoxia, and at least prevent organ failure [5]. Central venous oxygen saturation (ScvO 2 ) and serum lactate concentration are commonly accepted indicators of oxygen supply and global oxygen balance [3,6,7] and have been suggested as goals of treatment for more than a decade. Serum lactate has traditionally been interpreted as a marker of tissue hypoxia due to inadequate oxygen delivery. Particularly during prolonged episodes of sepsis, however, high levels of serum lactate can also result from impaired hepatic clearance or increased pyruvate production which is irresponsive to an optimized oxygen supply [8]. By now there are doubts concerning treatment strategies which are focused on serum lactate levels [9].
A level of ScvO 2 above 70% has been a conventional goal in early septic shock therapy since first early goal directed therapy (EGDT) studies have been released [10,11]. Recent trials failed to demonstrate a survival benefit for these classic treatment goals compared to usual care [12][13][14][15]. Particularly in later stages of disease, interpretation of ScvO 2 is challenging for clinicians [7]. Reviews attest "no benefit of early resuscitation guided by venous oxygen saturation" [16] or suggest that a subset of patients could benefit [15]. Consequently, adjusting ScvO 2 is no longer part of the current surviving sepsis campaign (SSC) guidelines [4]. In contrast, the SSC guidelines state that the use of the previous targets is still safe and may be considered as no harm was associated with the interventional strategies and clinicians still need guidance for approaching this group of patients with significant mortality and morbidity. This raises the question whether ScvO 2 should still be measured in septic shock at all.
ScvO 2 values are considered to reflect mixed venous oxygen saturation (SvO 2 ) [11]. SvO 2 is an indirect index of global oxygen supply-to-demand ratio [11]. In early septic shock with inadequate oxygen delivery, mostly due to anaemia, absolute or relative hypovolemia, or myocardial dysfunction, systemic extraction ratio rises and SvO 2 decreases [17]. In prolonged septic shock oxygen utilization is compromised by oedema induced diffusion limitation and by toxic respiratory chain uncoupling [5,18,19]. As a result less oxygen is consumed and returned back to the central circulation as indicated by increasing SvO 2 values. Increased risk of multiple organ dysfunction and death, hence, are likely reflected by both low and high ScvO 2 values. This suggests there may be ranges or corridors of ScvO 2 values, which are associated with favourable patient outcome. Moreover, the significance of specific ScvO 2 values may vary in the course of septic shock and thus a single fixed goal or corridor may be inadequate for treatment guidance. Data supporting these hypotheses are scarce, retrospective analyses suggest reduced survival for patients with venous saturations below 70% and above 90% in the first 6 h [19], as well as above 80% up to day three of septic shock therapy [20]. Prognostic studies evaluating ScvO 2 with a time-dependent approach, however, are lacking.
The aim of this study therefore was to link the level of ScvO 2 with mortality rates and identify ScvO 2 levels in defined time corridors potentially beneficial for patient outcome from septic shock.

Materials and Methods
This investigation is single centre retrospective observational, using the patient data management system (PDMS) database (IntelliVue Clinical Information Portfolio, Phillips Healthcare, Netherlands) of a surgical intensive care unit (ICU). The study is designed corresponding to the declaration of Helsinki and ethical approval has been granted by a local committee (Medical Faculty of Mannheim, Germany, chairman Prof. JP Striebel, No. 2014-816R-MA). Design and reporting are according to the CONSORT statement recommendations (www.consort-statement.org).

Cohort inclusion and exclusion criteria
All patients admitted with septic shock between 1 st January 2007 and 31 st October 2013 were included ( Figure 1). Septic shock was operationalized according to the SSC guidelines of 2012 focussing on sepsis-induced hypotension despite appropriate volume resuscitation [3]. Patients admitted to the ICU with an International Classification of Disease 10-code implying sepsis who had received norepinephrine and at least one measurement of ScvO 2 within 72 h after ICU admission were identified automatically from the PDMS. Afterwards all matching electronic records as well as additional clinical data were examined by a medical expert. Clinical records of previous ICU stays were also reviewed. Patients who were readmitted due to ongoing septic disease were excluded as well as cases in which septic shock developed after admission to the ICU. Patients treated with norepinephrine for less than 3 h after having received a fluid bolus, whose venous catheter was not centrally located, or who have been receiving extracorporeal gas exchange therapy within their first 72 h were also excluded.

Cohort characteristics and outcome
Of all patients meeting the cohort of septic shock basic demographic data, sequential organ failure score (SOFA) and simplified acute physiology score 2 (SAPS 2) [21,22] were extracted from the PDMS. ScvO 2 measurements as well as hemodynamic monitoring parameters and routine laboratory parameters have been determined five times -first 6,7 to 12,13 to 24,25 to 48 and 49 to 72 h after ICU admission. If more than one value was available, the one closer to the upper time interval limit was considered with exception of the first interval in which first measurements were considered. This procedure allowed comparability to previous investigations [19,[23][24][25][26]. Based on recent literature [10,20] and our receiver operating characteristic (ROC) analysis, ScvO 2 levels of each time interval were categorized as below, within or above corridors of 66-75% and 71-80%. Furthermore, the minimum and maximum ScvO 2 values during the first 72 h were extracted.
Necessity of organ replacement therapy as well as ICU length of stay were also determined. Follow-up ended at hospital discharge or inhospital death, which were extracted from the local hospital information system (IS-H med, SAP-Industry Solution, Germany).

Note on clinical practice during the investigation
Moderate sedation, lung protective ventilator settings supporting spontaneous breathing, empirical antibiotics, early microbiological and radiological diagnostics as well as interventional and surgical focus control were ensured like recommended by the SSC [3]. For hemodynamic stabilization balanced crystalloid solutions, norepinephrine and dobutamine were preferred choices. A MAP of>65 mmHg, urine output of >0.5 ml/kg/h, ScvO 2 >70% and serum lactate concentration<2 mmol/l were declared as primary goals. If required, hemodynamic monitoring was augmented by echocardiography or pulse contour cardiac output (therapeutic aim: cardiac index (CI)>2.5 l/min * m 2 ) and to estimate volume status and contractility. ScvO 2 measurements were part of routine care, determined by a point-of-care blood gas analyzer (ABL800 Radiometer, Copenhagen, Denmark) and directly transmitted to the local PDMS.

Statistical analyses
Descriptive statistics were used to summarize demographic, clinical, hemodynamic and laboratory parameters of the cohort and also for survivors and non-survivors separately. Characteristics of the latter two groups were compared with two-tailed t-tests and chi-square tests. Survivors and non-survivors of the cohort were classified according to their ScvO 2 -level at each time interval. For each of the fivetime intervals, odds ratios (OR) with 95% confidence intervals (CI) for ScvO 2 -dependent mortality risk in those still alive at the beginning of the respective interval were obtained from logistic regression. ROCcurves were used to determine the maximum value for the sum of For all analyses a p-value<0.05 (two-tailed) was considered statistically significant. SAS 9.4 (SAS Cooperation, Cary, North Carolina, USA) was used for data analysis.

Results
Of 16044 ICU admissions during the study period 466 potential shock patients fulfilled the inclusion criteria. After examination by an expert the final cohort of 238 patients with septic shock on admission was set ( Figure 1). Overall hospital mortality of 57.6% and observed disease severity of the first day after ICU admission based on SAPS 2 and SOFA score in these patients are consistent with manifest multiple organ failure (Table 1). Survivors had higher haemoglobin levels at day one (p<0.05) and lower plasma lactate levels in any time interval (p<0.05 for each interval). A nonsignificant trend to higher initial ScvO 2 measurement values in survivors compared to non-survivors (mean ± SD: 75.7% ± 9.1% vs. 73.1% ± 10.1%, p=0.13; median: 77% vs. 74%) and lower ones on day three (71.2% ± 7.7% vs. 73.1% ± 11.6%, p=0.37; median: 72% vs. 75%) was detected ( Figure 2).    Central venous oxygen saturation; * significant result by two-tailed ttest comparison.
The minimum value of ScvO 2 was 69.03% on average in all patients (range 35.2%-95.3%). While ScvO 2 minima did not differ in survivors and non-survivors, maximum ScvO 2 values, on average, were significantly higher in survivors compared to non-survivors ( Figure  3a). Hospital mortality was higher if maximal ScvO 2 values recorded at any time within the first three days of therapy were below 70%, whereas no impact on outcome was detected for maximum values ≥ 80% (Figure 3b).  On day three ScvO 2 cut-offs above 70% distinguished well between survival until patient's discharge from hospital and intra-hospital mortality (Figure 4). A ROC analysis detected an optimal cut-off at 76.0% for third day's ScvO 2 (70.4% sensitivity, 60.6% specificity; area under the curve: 0.662). ScvO 2 ≥ 84.0% (n=6) on day three identified non-survivors with a sensitivity of 100%. Figure 4: In-hospital mortality in dependence of variable ScvO 2 cutoffs. Comparative representation of in-hospital mortality in dependence upon ScvO 2 cut-off values over time course of ICU stay-patients with ScvO 2 values below the respective limit are shown in deep blue, those above the limit in light blue; ScvO 2 , central venous oxygen saturation; *Significant result from chisquare test.
Odds ratios for in-hospital mortality of patients classified in defined ScvO 2 corridors (66-75% and 71-80%) did not indicate any difference regarding mortality risk for patients with ScvO 2 values below or above these reference limits at the times of measuring within 48 h after ICU admission ( Figure 5). Mortality was higher in patients with ScvO 2 values>75% (n=32) compared to those in the corridor of 66-75% (n=28) on day three with an odds ratio of 3.6 (95% confidence interval

Discussion
Based on early EGDT studies [10,27] former treatment recommendations supported ScvO 2 correction above a value of 70% within the first 6 h of sepsis [3,15]. In accordance Pope et al. [19] reported mortality rates to be significantly elevated if septic patients in the emergency department indicated maximum ScvO 2 values below 70% (hypoxic range) compared to those being in normoxic range (ScvO 2 71-89%). Park et al. confirmed these findings for septic shock patients exhibiting oxygen extraction ratios higher than 0.3, being approximately equivalent to calculated ScvO 2 values lower than 70% [28] and Lee et al. also found reduced survival if ScvO 2 could not be raised above 70% and lactate remained abnormal [29]. In line with these observations, our data showed diminished survival rates if ScvO 2 values could not be increased above this threshold (<70%) during the first 72 h of ICU stay. About one third of our patients have had ScvO 2 values below 70% and approximately in one fifth ScvO 2 was below 65% when first measured in ICU. These results are supported by findings of Boulain et al. who prospectively showed that ScvO 2 values<70% are frequent (27%) at ICU admission [30] and extended by van Beest et al., who demonstrated that even lower rates of ScvO 2 (<60%) are still occurring (6%) [25]. Initial mean ScvO 2 values (74% ± 11% [25] and 74 ± 10.4% [30]) were comparable to ours and there was also no difference detected between survivors and non-survivors [30]. On the other hand Park et al. [28] as well as Pope et al. [19] reported significantly elevated in-hospital mortality rates if ScvO 2 values and oxygen extraction rates were in the hyperoxic range, i.e., above 89% or lower than 20%, respectively, with the latter most likely indicating ScvO 2 values being above 80%. These observations were made within the first hour and after 6 h of ongoing goal directed therapy. The data reported on by Textoris et al. [20] confirmed and extended the period of increased mortality risk in septic patients by investigating maximum ScvO 2 values taken during 72 h of treatment, although this group gave no information on the exact time of ScvO 2 determination: Using a ScvO 2 cut-off value of 80%, a 1.5 fold higher mortality rate for patients with maximum values above 80% was shown compared to those with values below this threshold (48% vs. 30%, p<0.05).
Our analysis of mortality rates associated with ScvO 2 in the five time intervals using different cut-off values for ScvO 2 starting from 65% up to 85% revealed significantly higher mortality rates with ScvO 2 values above 75% on the third day. ROC analysis showed an optimal cut-off at 76.0% for third day's ScvO 2 (70.4% sensitivity, 60.6% specificity). Moreover, mortality was significantly higher in patients with ScvO 2 >75% (n=32) compared to those in the corridor of 66-75% (n=28) on day three with an odds ratio of 3.6 (confidence interval 1.2 to 10.8, p=0.02) ( Figure 5). Kaplan-Meier analysis of mentioned ScvO 2 corridor on day three confirmed lower survival rates for ScvO 2 values below and above the corridor of 66-75%, reaching the level of statistical significance for ScvO 2 values above 75% ( Figure 6). Thus, patients whose ScvO 2 values were in the corridor of 66-75% on third day of septic shock had threefold higher odds of survival compared to those whose ScvO 2 was above 75%. This is a drastically lower limit than previously suggested (Pope et al. >90% [19], Textoris et al. >80% [20]) and within the previously recommended range of 70-80% [3,11]. Moreover, these recommendations have been extrapolated to nearly every kind of critical illness, perioperative situations [16,31] and decision-making concerning transfusion [32,33]. According to the present study ScvO 2 ranges above 75% have to be considered as potentially deleterious for patients in later stages of septic shock. This is supported by the observation of no single individual, whose ScvO 2 was above 85% on the third day of septic shock surviving. Unsurprisingly, specificity of ScvO 2 values ≥ 84.0% on day three regarding in-hospital mortality was 100%.
Overall the most important finding was a noticeably increased mortality risk even if ScvO 2 values were only slightly elevated (>75%) in later stages of septic shock, which until now have been considered as unsuspicious. They might be used as warning signs as recommended by some authors in the perioperative cardiosurgical setting [24,34].

Limitations
This is a retrospective study. Therefore the observed associations may not be causal and the results have to be confirmed by adequately powered prospective studies. Because the data are from a single center, particularities of local practice may have had an impact on measurement frequency, but also magnitude of ScvO 2 values. In addition, ScvO 2 measurements were determined for routine care leading to factors of uncertainty such as mix-up of patients and documentation errors. Considering predefined time corridors retrospectively leads to data gaps in some patients because measurements are taken in irregular, clinically necessary and practical intervals. In contrast, for other patients only a fraction of available measurements was used. Incomplete data also result from patients dying during the observation period. Nonetheless we analysed numerous ScvO 2 values from septic patients with high mortality risk. In spite of these limitations, we consider it important to disseminate these results because they indicate that potentially harmful ScvO 2 levels are apparently very close to formerly recommended treatment goals. In future prospective studies ScvO 2 should ideally be measured regularly if not continuously in all patients irrespective of their clinical course. Specifically, in such studies the value of ScvO 2 for mortality prediction by U-shaped models should be addressed in a time course dependent fashion.

Conclusion
We strengthened the hypothesis that ScvO 2 is of potential value in predicting unfavourable outcome in septic shock. If ScvO 2 values of at least 70% have not been achieved within 72 h of therapy, in-hospital mortality significantly increased. On the other hand ScvO 2 values above 75% on day three were associated with a threefold mortality rate compared to values in a ScvO 2 range of 66-75%. High values in early septic shock seem to be unsuspicious as well as lower ones in later stages. Further investigations have to confirm these findings in before they can be implemented in clinical decision-making.