Oxygen Delivery and Oxygen Consumption in Pediatric Fluid Refractory Septic Shock During the First 42 h of Therapy and Their Relationship to 28-Day Outcome

Background: In septic shock, both oxygen delivery (DO2) and oxygen consumption (VO2) are dysfunctional. The current therapeutic regimens are geared to normalize global oxygen delivery (DO2) to tissues via goal directed therapies but mortality remains high at 10–20%. Methods: We studied cardiac index (CI), systemic vascular resistance index (SVRI), central venous oxygen saturation (ScvO2), central venous pressure (CVP), peripheral oxygen saturation (SpO2), mean blood pressure (MBP), body temperature, blood lactate, base excess and hemoglobin concentration (Hb) in a cohort of children admitted in “fluid-refractory” severe septic shock to pediatric intensive care, over 4.5-years. We calculated their 6 h global oxygen delivery (DO2) and global oxygen consumption (VO2) over the first 42 h and looked at factors associated with VO2/DO2 ratio (i.e., global oxygen extraction, gO2ER) and 28-day mortality. Results: Sixty-two children mean age (SD) 7.19 (5.44) years were studied. Fifty-seven (93%) children were sedated and mechanically ventilated and all received adrenaline or noradrenaline or both and added milrinone in 6 (9.6%). At 28 days, 9 (14.5%) were dead. The global oxygen extraction ratio (gO2ER) was consistently lower amongst the survivors and independently predicted mortality (ROC AUC = 0.75). A lactate level of 4 mmol/l or above, when associated with a concurrent metabolic acidosis predicted mortality with a sensitivity of 100% (95% CI 90.5–100) and a specificity of 67.7% (95% CI 62.2–72.9). A gO2ER of 0.48 or above on admission to the PICU was associated with death with a 66.7% sensitivity (95%CI 29.9–92.5) and 90.5% specificity (95%CI 79.3–96.8). A global O2ER of >0.48 combined with a concurrent blood lactate >4.0 mmol/l at any time within the first 42 h of therapy predicted death with a sensitivity of 63.9% (95% CI, 46.2–79.1) and specificity of 97.8% (95% CI, 95.7–99.0). A radar plot identified MBP-CVP difference, and CI as additional goals of therapy that may offer a survival benefit. Conclusions: Global O2ER of >0.48 with a concurrent blood lactate >4.0 mmol/l in children with metabolic acidosis was an independent factor associated with death in fluid resistant septic shock. Trends of gO2ER seem useful to recognize survivors and non-survivors early in the illness.

Background: In septic shock, both oxygen delivery (DO 2 ) and oxygen consumption (VO 2 ) are dysfunctional. The current therapeutic regimens are geared to normalize global oxygen delivery (DO 2 ) to tissues via goal directed therapies but mortality remains high at 10-20%.
Methods: We studied cardiac index (CI), systemic vascular resistance index (SVRI), central venous oxygen saturation (ScvO2), central venous pressure (CVP), peripheral oxygen saturation (SpO2), mean blood pressure (MBP), body temperature, blood lactate, base excess and hemoglobin concentration (Hb) in a cohort of children admitted in "fluid-refractory" severe septic shock to pediatric intensive care, over 4.5-years. We calculated their 6 h global oxygen delivery (DO2) and global oxygen consumption (VO2) over the first 42 h and looked at factors associated with VO2/DO2 ratio (i.e., global oxygen extraction, gO2ER) and 28-day mortality.
Results: Sixty-two children mean age (SD) 7.19 (5.44) years were studied. Fifty-seven (93%) children were sedated and mechanically ventilated and all received adrenaline or noradrenaline or both and added milrinone in 6 (9.6%). At 28 days, 9 (14.5%) were dead. The global oxygen extraction ratio (gO2ER) was consistently lower amongst the survivors and independently predicted mortality (ROC AUC = 0.75). A lactate level of 4 mmol/l or above, when associated with a concurrent metabolic acidosis predicted mortality with a sensitivity of 100% (95% CI 90.5-100) and a specificity of 67.7% (95% CI 62. 2-72.9). A gO2ER of 0.48 or above on admission to the PICU was associated with death with a 66.7% sensitivity (95%CI 29.9-92.5) and 90.5% specificity (95%CI 79. 3-96.8). A global O2ER of >0.48 combined with a concurrent blood lactate >4.0 mmol/l at any time within the first 42 h of therapy predicted death with a sensitivity of 63.9% (95% CI, 46.2-79.1) and specificity of 97.8% (95% CI, 95.7-99.0). A radar plot identified MBP-CVP difference, and CI as additional goals of therapy that may offer a survival benefit.
Conclusions: Global O2ER of >0.48 with a concurrent blood lactate >4.0 mmol/l in children with metabolic acidosis was an independent factor associated with death in fluid resistant septic shock. Trends of gO2ER seem useful to recognize survivors and non-survivors early in the illness.
Keywords: septic shock, children, oxygen delivery, oxygen consumption, mortality BACKGROUND Childhood mortality in invasive infections, sepsis and septic shock remains high at 3.9, 5.6, and 17.0%, respectively (1,2). Significant advances in the management of septic shock in children have been made and it hails from standardized, timely, goal directed therapies to normalize "macro-circulation" abnormalities (3). Clinicians target global oxygen delivery (DO 2 ) as the therapeutic goal with the hope of matching the supplydemand gap of oxygen during the treatment of shock.
In septic shock, both oxygen delivery and oxygen consumption are dysfunctional. Some authors claim normalizing ScvO 2 as an additional goal of therapy may be beneficial (4) whilst others argue that higher than normal ScvO 2 may increase the mortality risk (5,6). Higher ScvO 2 can indicate improved oxygen delivery but it does not rule out continuing hypoxia (7). This is because high ScvO 2 can also reflect poor tissue oxygen consumption resulting from sepsis induced dysfunctional capillary perfusion and altered mitochondrial function. However, quantitative estimations of these are scant.
Microcirculatory alterations in sepsis are associated with organ dysfunction and mortality (8,9). Due to practical difficulties of measuring regional oxygen consumption and oxygen delivery, most clinical studies have focused on global oxygen delivery (DO 2 ) and global oxygen consumption (VO 2 ). Oxygen delivery is formally calculated using the degree of hemoglobin (Hb) oxygen saturation, and dissolved O 2 content in arterial blood and cardiac output (CO). The oxygen consumption (VO 2 ) is a composite estimate of global oxygen utilization. It is the calculated oxygen content difference between arterial and venous blood standardized to cardiac index.
To understand the evolution of DO 2 and VO 2 in pediatric "fluid resistant" septic shock during the resuscitation and stabilization phase and its associations, we estimated global oxygen delivery (DO 2 ) and global oxygen consumption (VO 2 ) and outcomes in a cohort of children in "fluid resistant" septic shock who were receiving standard goal directed therapy in the pediatric intensive care.

METHODS
We retrospectively studied hemodynamic variables during initial stabilization for 42 h and 28-day outcome in all children admitted from June 2009-Feb 2014 in fluid refractory septic shock to a regional 16-bed Pediatric Intensive Care Unit (PICU). Shock was classified as "fluid-refractory" when the total fluid requirement during resuscitation exceeded 60 ml/kg fulfilling international consensus conference criteria (10, 11) and required inotrope/vasopressor/vasodilator therapy.
All children had continuous hemodynamic monitoring including invasive arterial blood pressure and central venous pressure (via SVC catheter) as standard. Vasoactive agents (noradrenaline, adrenaline or milrinone) were used to optimize hemodynamics as per the ACCM-CPP protocol (12). In addition, their cardiac index (CI) and systemic vascular resistance index (SVRI) were measured approximately 6-h by an Ultrasonic Cardiac Output Monitor (USCOM model 1-A, USCOM Limited Australia http://www.uscom.com.au/) using a trans-cutaneous Doppler probe placed at the supra sternal notch (13,14).
We calculated global oxygen delivery (DO 2 ) using a standard formula (Equation 1) using pulse oximetry (SpO 2 ) as an suitable surrogate for oxygen saturation of arterial blood (SaO 2 ) (15). DO 2 was standardized to body surface area (DO 2 I).
We also obtained simultaneous measured central venous oxygen saturations (ScvO 2 ) (16-18) of venous samples obtained via radiologically confirmed central venous catheters placed in superior vena cava (SVC) and estimated mixed venous oxygen content. Global oxygen consumption (VO 2 ) was derived by subtracting estimated oxygen content in mixed venous blood from that of arterial blood (Equation 2). This too was standardized to body surface area (VO 2 I). We used the ratio of global oxygen consumption (VO 2 ) to global oxygen delivery (DO 2 ) standardized to body surface area i.e., VO 2 I/DO 2 I ratio (global oxygen extraction ratio -gO 2 ER, normal 0.28) as a crude clinical measure of global "efficacy" of oxygen consumption during resuscitation and stabilization phases of sepsis. We contemplated that global "efficacy" depended upon a combination of factors including the ability of tissues to extract oxygen and the reserve capacity of the cardiovascular system to increase oxygen delivery to match the peripheral oxygen demand. A higher gO 2 ER value was therefore suggestive of inadequate oxygen delivery (DO 2 ) or increased oxygen consumption (VO 2 ) or both. A lower gO 2 ER was reflective of increased oxygen delivery or decreased oxygen consumption (VO 2 ) or both.

Statistics
Absolute values were standardized to body surface area to adjust for age as appropriate. Continuous variables were summarized as mean (SD) and categorical data as count (percentage). Student t-test was used to test differences in continuous variables and the χ 2 test to compare proportions. All analyses were performed with statistical software IBM-SPSS version 20 (IBM Corporation, New York, USA). All tests were 2-tailed, and P < 0.05 was considered statistically significant.
Using gO 2 ER values, a receiver operating characteristic curves (ROC) was developed against outcome. The area under the curve (AUC) of more than 0.75 was considered clinically useful (19).
To have an "aerial view, " we also created radar plots of the above parameters observed over the first 42 h for survivors and non-survivors separately. These plots represented mean ± SE of each variable to enable a visual comparison between surviving and non-surviving populations.

RESULTS
There were 62 children (M: F = 34:28) aged mean (SD) 7.19 (5.44) years admitted in fluid refractory septic shock. Nine required continuous veno-venous hemofiltration (CVVH) during ICU stay. Upon admission to the PICU, 93.5% children were mechanically ventilated and sedated using a combination of morphine and midazolam. At 42 h 85% of the surviving children were still on ventilators. A demographic difference between survivors and non-survivors was not observed but the PIM2 score was worse as expected in non-survivors ( Table 1).
Their progression of CI, SVRI and the use of inotropes and vasopressors are shown in Figures 1A,B and   A gO 2 ER of 0.48 or above on admission to the PICU predicted death with a 66.7% sensitivity (95%CI 29.9-92.5) and 90.5% specificity (95%CI 79.3-96.8). The ROC curve of blood lactate levels of all children observed during the first 42 h at 6 hourly intervals against 28-day mortality also showed an AUC of 0.97. A lactate level of 4mmol/l or above, associated with a concurrent metabolic acidosis within the first 42 h of therapy linked to mortality with a sensitivity of 100% (95% CI 90.5-100) and a specificity of 67.7% (95% CI 62.2-72.9). However, the base excess alone was nonpredictive of the mortality risk (ROC AUC was 0.25) (see Figure 4).  Oxygen supply dependence graphs were generated for survivors and non-survivors (Figure 5). A critical DO 2 was not demonstrable in these curves. The SpO 2 -ScvO 2 gap had a significant negative correlation with CI (r = −0.16, p = 0.001) and SVRI (r = −0.28, p < 0.001) in survivors. However, SpO 2 -ScvO 2 gap neither correlated with CI (r = −0.07, p = 0.5) nor with SVRI (r = −0.17, p = 0.2) in non-survivors.   VO 2 I significantly positively correlated (2 tailed, Pearson) with CI (r = +0.817, p < 0.001) and negatively with SVRI (r = −0.542, P < 0.001). The overview radar plots of above clinical parameters drawn separately for survivors and non-survivors with the respective ± 2SE fills enabled visual recognition of the areas of disparity that needs attention to bring the non-survivor plot closer to that of survivors (Figure 6).
Neither the Hb on admission nor the body temperature was associated with outcome. The O 2 ER correlated with blood lactate levels (Figure 7).

DISCUSSION
We have presented a comprehensive account of the cardiovascular parameters in a cohort of children in "fluid  resistant" septic shock during the first 42 h of bundle therapy in intensive care. Their severity of illness is indicated by the PIM2 scores (20). The titration of inotropes and vasopressors during treatment was assisted by non-invasive 6 hourly cardiac output monitoring. The limitation of our study to 42 h was based on lack of data resulting from steadily receding need for inotropes (21) and clinicians deciding to taper off invasive monitoring and regular cardiac output measurements in children showing improvement. We also looked at the physiological variables that accounts for DO 2 and its influence upon mortality and the reliability of ScvO 2 as a therapeutic efficacy evaluation tool.

Critical DO 2
The expected normal values for global DO 2 I and global VO 2 I are 500 ml /min/m 2 , and 150 ml/min/m 2 respectively in adults (22). In the new born DO 2 I is high at birth reducing to a mean (SD) of 609 (141) ml/min/m 2 at 3 days of age (23). We noted that global DO 2 I above 600 ml/min/m 2 during the first 42 h of  therapy in PICU was more likely to be associated with survival (Figure 2) in our cohort. A trial involving adult high-risk surgical postoperative patients too has identified a global DO 2 I above 600 ml/ min/m 2 as critical for survival (24).
We could not demonstrate a critical DO 2 in our cohort, especially non-survivors (Figure 5). This refers to the point at which the compensatory increase in O 2 extraction is exhausted with the lowering DO 2 , and the VO 2 becomes dependent on DO 2 (25). This pathologic O 2 supply dependency possibly results from loss of efficient tissue distribution of DO 2 (26). In fact, this abnormal dependence of VO 2 on DO 2 and the absence of a critical DO 2 , may be the "hallmark" of fluid refractory septic shock state. In other words, tissue oxygen consumption is not met by the delivered oxygen despite maximum extraction leading to a state of chronic tissue oxygen debt. The longer this persists; more likely the shock state becomes irreversible as we have noted in our gO 2 ER plot of non-survivors (Figure 3). In adults too, VO 2 dependency on DO 2 is considered a marker of septic shock (27). It seems that clinically deliverable ranges of global DO2 achieved via current bundle therapies were insufficient to rectify this state of VO 2 dependence on DO 2 in our non-survivors. This observation is in agreement with that of recent epidemiological data showing children who die early from refractory septic shock, usually die of cardiovascular demise (28).

gO 2 ER
Upon admission to PICU, gO 2 ER was approximately 0.5 in most children (Figure 3). Thus, gO 2 ER was less sensitive in predicting mortality on admission. However, we noted a steady improvement in gO 2 ER toward normal levels (0.25-0.3) with a narrow 95% CI during the first 42 h of therapy along with a progressive reduction in global VO 2 I in survivors (Figure 2). In contrast, a persistently high mean gO 2 ER was observed in nonsurvivors (Figure 3) with a progressively widening 95% CI due to reduced number of data following the demise of some children. A gO 2 ER greater than 0.5 at 6 h was a recognized risk factor of death after cardiac operations in infants (29). We also observed that the first 42-h gO 2 ER value was independently associated with 28-day mortality (Figure 3).
In our cohort, a high gO 2 ER resulted from a disproportionately lower DO 2 due to a lower CI especially amongst non-survivors (Figure 2). This might reflect an inability of the myocardium to respond to the demand i.e., a state of myocardial "stunning" alluded to mitochondrial dysfunction (30)(31)(32). Unmasking of myocardial weakness during treatment of septic shock is known (21). We observed that persistently high gO 2 ER and lack of attainment of critical DO 2 in the initial phases of therapy may identify children in septic shock with chronic tissue oxygen debt progressing to an irreversible shock state and death. This suggests that early options of treatment above the standard is needed in the above sub-set of children, for example, inodilators such as milrinone or levosimendan, or even extracorporeal membrane oxygenation (ECMO) to attain critical DO 2 and save life (33) (34). This is because our data suggest that the DO 2 was far short of matching the oxygen need or consumption especially amongst non-survivors. In this context, gO 2 ER seems a valid and independent global estimate of the real time "efficacy" of O 2 consumption in this clinical scenario and a tool to identify children at high risk of death although cellular oxygen uptake and utilization are not verifiable from global measurements (35).

ScvO 2
When DO 2 is insufficient to meet metabolic requirements, increased tissue oxygen extraction reduces oxygen content in effluent venous blood (16).Thus, a high SpO 2 -ScvO 2 gap is a crude estimate of high oxygen consumption at tissue level. This explains its crude use as a marker of inadequate tissue perfusion and a therapeutic goal in critical care practice. We observed a strong correlation between SpO 2 -ScvO 2 gap and gO 2 ER as expected.
It is also contemplated that ScvO 2 may read higher when O 2 consumption is reduced at capillary level (16)(17)(18). In other words, SpO 2 -ScvO 2 gap should be lower in the presence of arteriovenous (A-V) shunting i.e., in severe sepsis. On the contrary, we observed that SpO 2 -ScvO 2 gap significantly correlated negatively with CI and SVRI in survivors, i.e., with clinical improvement. But, it neither correlated with CI nor with SVRI in non-survivors. This suggested that ScvO 2 had little monitoring value in children in whom it matters most i.e., in non-survivors. Although ScvO 2 has been shown to be useful in guiding resuscitation (36), its validity in the resuscitated subjects remains a matter of controversy (37). Further, a reduction in ScVO 2 when CI was < 3.3 as well as > 6.0 (l/m 2 /min) has been observed in pediatric septic shock (21). This may also explain why we cannot overcome flow dependence as CI increases in our cohort (Figure 5).

Lactate
Although lactate and base excess correlate, it's different between survivors and non-survivors (Figure 4). High lactate on hospital admission, and at 12 and 24 h in pediatric septic shock is reported to predict death (38). It appears that the risk of death was highest only when a metabolic acidosis (as detected by base excess) was present concurrently with a blood lactate level > 4 mmol (39). This is possibly because there were many other factors in addition to poor tissue perfusion contributing to the rise in blood lactate; for example, the use of adrenaline, renal impairment and liver dysfunction (40). This affirms that clinical and prognostic importance of an elevated lactate differs widely by disease state (39). In our study, gO 2 ER correlated positively with blood lactate (Figure 7), but lower levels of lactate amongst some non-survivors reduced its sensitivity in recognizing high mortality risk (Figure 7). This explains why a global O 2 ER of >0.48 combined with a concurrent blood lactate >4.0 mmol/l and a metabolic acidosis enhanced its sensitivity of assessing risk of death. The high O 2 ER together with metabolic acidosis makes the origin of lactate more likely from tissue anaerobic metabolism. Therefore, despite the confounding factors, we infer that gO 2 ER in combination with lactate can predict mortality with a significant sensitivity and specificity. However, negative base excess alone had no value in mortality prediction. Data to support or refute treatment of lactate acidosis with a buffer (sodium bicarbonate) or its clearance through dialysis in this scenario are scant (41).

Radar Plots
Our overview radar plot of cardiovascular variables was an attempt to recognize all vital parameters and therapeutic monitoring targets between non-survivors and survivors in a single figure (Figure 6). The use of mean ± 2 SE fills was an attempt to recognize what parameters may be significantly apart between the two populations i.e., surviving and non-surviving children presenting in fluid refractory septic shock. We presume that closing the gap in these parameters toward that of the survivors may improve chances of survival. The radar plot suggested areas of therapy that may be optimized. For example, increasing mean MBP-CVP gap by reducing the upper limit of CVP to 8 cm H 2 O, enhancing the lower limit of SVRI and increasing the cardiac index (by mechanical means) could be lifesaving if achieved concurrently. We have learned that fluids in excess of 40 mls/kg in resuscitation may improve survival in septic shock in children (42) but, overzealous use may be detrimental (43). A CVP of 8 cm H 2 O or below has been linked to lowest mortality in septic shock (43).

Limitations of this Study
The retrospective nature of our study limits its value, but all data used in this study has been prospectively collected in real-time and validated (i.e., authenticated and saved by the supervising nurse daily) and archived electronically in the "MetaVision" (Clinical Information System) database of the PICU. Thus, inherent errors of retrospective clinical studies based on paper records were minimized.
Another limitation of our study was ScvO 2 . Historically, site dependent measurement errors have rendered ScvO 2 less valuable. Since ScvO 2 can vary due to preferential blood flow patterns of superior vena cava (SVC) and inferior vena cava (IVC) (17,44) and admixture of highly desaturated coronary blood flow, the ScvO 2 may not reflect true oxygen consumption. In children with septic shock, low CI is generally associated with lower ScvO 2 ; however, ScvO 2 may also remain less than 70% in normal and in high cardiac output states (e.g., in warm shock) (21).
Similarly, the use of peripheral perfusion dependent SpO 2 as a surrogate marker of SaO 2 could have underestimated the DO 2 I especially in cold shock states. However, the margin of error ensued is likely to be very small as our mean (SD) SpO 2 was 96.1 (3.6). On the other hand, the use of ScvO 2 in the calculation of VO 2 I may also have added a margin of error. ScvO 2 is more variable than SvO 2 and tends to be an over-estimate in critically ill patients (45). It is also observed to remain less than 70% in some children with normal and high CI in septic shock (21). Although ScvO 2 is useful in guiding resuscitation (36), its validity in the resuscitated is controversial (37).
CI measurement by USCOM is also subject to wide variation. In order to minimize inter-observer variability, USCOM operators were certified competent by the author (AD) as advocated by previous researchers (46).
In severe sepsis states, changes in ScvO 2 due to preferential blood flow pattern and microcirculatory shunting at tissue level are both contributors to gO 2 ER. Therefore, the persistently high gO 2 ER that we observe in this study as a predictor of death is a global crude reflector of not just oxygen consumption, but also inherent circulation abnormalities associated with severe sepsis including microcirculatory shunting.
A drawback is that gO2ER cannot be used worldwide as CI is not measured at all PICUs and CI measurements have its own limitations (47,48). However, our findings support its use in very specific conditions such as septic shock. This is in line with surviving sepsis guidelines emphasizing the early use of age-specific therapies to attain time-sensitive goals, specifically in the first hour of fluid resuscitation and inotrope therapy, and subsequent intensive care unit hemodynamic support (12).

CONCLUSION
We observe that absolute ScvO 2 , and base excess were less robust indicators of outcome during initial 42 h of therapy of "fluid resistant" pediatric septic shock. In contrast, global O 2 ER of >0.48 found concurrently with a blood lactate >4.0 mmol/l and metabolic acidosis seems a better independent predictor of death.
We conclude that trends of gO 2 ER may recognize survivors and non-survivors early in the illness. Although, further studies are required to assess the validity of this bedside measure, a nonimproving gO 2 ER may justify implementation of extraordinary measures of therapy such as mechanical cardiac support and novel pharmacologic strategies early in pediatric septic shock.

ETHICS STATEMENT
This was an observational study of routinely monitored clinical parameters in septic shock in the PICU. No patient identifiable data was stored. Therefore, according to the local guidance the study was registered as a service evaluation project at King's College Hospital, London (Clinical Audit Support System (CASS) project no. 2902).

AUTHOR CONTRIBUTIONS
All authors contributed to manuscript revision, read and approved the submitted version. AD led collection of data; CG performed the statistical analysis, prepared figures and tables and wrote the first draft of the manuscript and managed revisions. JC assisted interpretation of data in the clinical context and previous research findings.

FUNDING
Open access publication support from USCOM Australia Ltd.