Introduction

Hemodynamic optimization through fluid resuscitation is commonly used in critically ill patients with tissue hypoperfusion. The goal of volume expansion (VE) is to raise cardiac output and oxygen supply to improve tissue oxygenation. Recognizing patients who would benefit from VE remains challenging1. Identifying such patients is often dependent on measuring cardiac output2,3. Echocardiography is a skill that has made strides but not fully penetrated all critical care areas; it is limited by ultrasound availability and poor echogenicity, especially in mechanically ventilated patients. Passive leg raising test and end-expiratory occlusion methods also necessitate cardiac output measurement to asses fluid responsiveness4,5. Other measurements such as pulse pressure and stroke volume variations require specific technologies6 or are restricted to certain patient populations7. With these limitations, defining fluid responsiveness without cardiac output measurement would be of great help to the clinician at the bedside.

Venous-to-arterial CO2 tension difference reflects the balance between CO2 production and CO2 delivery to the lungs, a surrogate of the cardiac output8,9. Opposing changes over time in central venous-to-arterial CO2 tension difference (ΔPCO2) and cardiac output were reported in septic shock patients10,11,12. In post-cardiac surgery sedated and mechanically ventilated patients, Yazigi et al. observed a significant inverse correlation between changes in ΔPCO2 (Δ-ΔPCO2) and changes in cardiac index (ΔCI) induced by VE13. Moreover, in the same population, changes in central venous oxygen saturation (ΔScvO2) after VE were a reliable parameter to define fluid responsiveness14. However, data in the septic population is lacking.

In situations with tissue hypoxia, the increase in cardiac output and oxygen delivery (DO2) after VE would result in an increase in CO2 production (VCO2) and oxygen consumption (VO2) (oxygen supply dependency). These metabolic changes might confound fluctuations in arteriovenous O2 and CO2 parameters attributed solely to circulatory changes. This might reduce the changes in ΔPCO2 and ScvO2 induced by VE. Clinical studies have shown that the ratio of ΔPCO2 over the arterial-to venous oxygen content (ΔPCO2/ΔContO2) was a good indicator of oxygen supply dependency (tissue hypoxia) in critically ill patients15,16. This indicator could perhaps be used to identify such patients and guide the usage of CO2 and oxygen gaps.

Therefore, our study aimed to investigate: (1) if Δ-ΔPCO2 and ΔScvO2 are reliable parameters to identify fluid responsiveness in overall sedated and mechanically ventilated septic patients; (2) if the reliability of these parameters would be better in the sub-group of patients with no tissue hypoxia, defined as the absence of an increase in VO2 induced by a rise in DO2 after VE (oxygen supply independency); (3) if baseline ΔPCO2 /ΔContO2 ratio is a good predictor of tissue hypoxia. Such measurements are readily available in these patients with central venous and arterial catheters.

Materials and methods

This prospective and observational study was conducted in a single, mixed medical and surgical adult intensive care unit (ICU) between April and December 2017. The study was approved by the local institutional ethics committee (Comité d’Ethique du centre hospitalier du Dr. Shaffner de Lens). Informed consent was obtained from the next of kin of each patient. All experiments were performed in accordance with relevant guidelines and regulations.

Patients

We studied mechanically ventilated patients with sepsis17 for whom the attending physician decided to give VE due to the presence of at least one clinical sign of tissue hypoperfusion17 as previously described15: (a) systolic arterial pressure < 90 mmHg, mean arterial pressure < 65 mmHg, or the requirement for vasopressor administration; (b) skin mottling; (c) lactate levels > 2 mmo/l; or urinary output < 0.5 ml/kg/h for ≥ 2 h. Also, patients had to have a PiCCO device (PiCCO, Pulsion Medical System, Munich, Germany) as part of routine management of persistent signs of inadequate tissue perfusion in our ICU. Exclusion criteria were: pregnancy, age < 18 years old, moribund, and risk of fluid loading-induced pulmonary edema.

Measurements

Demographic data, acute circulatory failure etiology, the Simplified Acute Physiology Score (SAPS) II, and the Sequential Organ Failure Assessment (SOFA) scores were obtained on the day of enrollment. CI was obtained with the PiCCO device by central venous injections of 20 ml of iced 0.9% saline solution and recorded as the average of the three measurements.

Arterial and central venous blood gases were measured using the GEM Premier 4000 (Instrumentation Laboratory Co, Paris, France). The central venous blood was collected from a central venous catheter with the tip confirmed to be in the superior vena cava, near or at the right atrium, by radiograph as previously described15. ∆PCO2 was calculated as the difference between the central venous carbon dioxide tension and the arterial carbon dioxide tension. The arterial oxygen content was calculated as CaO2 (ml) = 1.34 × Hb (g/dl) × SaO2 + 0.003 × PaO2 (mmHg), where SaO2 is the oxygen saturation of arterial blood, Hb the hemoglobin concentration, and PaO2 the arterial oxygen tension. The central venous oxygen content was calculated as CcvO2 (ml) = 1.34 × Hb (g/dl) × ScvO2 + 0.003 × PcvO2 (mmHg), where PcvO2 is the central venous oxygen tension. ΔContO2 (ml) was calculated as CaO2 − CcvO2. DO2 (ml/min/m2) was calculated as CaO2 × CI × 10. VO2 (ml/m2) was calculated as CI × ΔContO2 × 10. Oxygen extraction was defined as OE = VO2/DO2. We also calculated the ∆PCO2/ΔContO2 ratio.

Study protocol

A first set of hemodynamic and oxygen-CO2 derived variables measurements was performed at baseline, including heart rate (HR), systemic arterial pressure, CI (thermodilution), DO2, VO2, ScvO2, arterial lactate level, and ∆PCO2. A 500 ml of colloid solution (4% Human serum albumin, Vialebex®; LFB) was administered to the patient over 15 min via a specific venous line. The same set of measurements was repeated immediately after the end of VE infusion. Ventilation parameters and infusions of norepinephrine and sedation drugs were remained unchanged during the VE.

Changes in hemodynamic and oxygenation variables were expressed as relative changes (([parameter after volume expansion − parameter before volume expansion]/parameter before volume expansion) × 100).

Statistical analysis

Patients in whom 500-ml VE increased thermodilution-derived CI > 10% were defined as responders and the remaining ones as non-responders. Also, patients were divided into two subgroups according to their increases in VO2 (≤ or > 10%) induced by VE. All data are expressed as mean ± SD, or as median [25–75%, interquartile range, (IQR)], as appropriate. The normality of data distribution was assessed using the Shapiro–Wilk test. Comparisons of values between responders and non-responders were performed by two-tailed Student’s t test, or Wilcoxon rank-sum test, as appropriate. Pairwise comparisons between different study times were assessed using paired Student’s t test or Wilcoxon signed-rank test, as appropriate. Analysis of categorical data was performed using the Chi2 and Fisher’s exact tests. Linear correlations were tested by using the Pearson or the Spearman test, as appropriate.

Receiver operating characteristic (ROC) curves were constructed to evaluate the ability of each parameter to predict fluid responsiveness after fluid challenge. The AUCs were compared using the nonparametric technique described by DeLong et al.18. Previously, we have shown that the upper 95% confidence interval values of the least significant changes (LSC), which are the minimum changes that needed to be measured by a laboratory analyzer in order to recognize a real change of measurement, for ΔPCO2 and ScvO2 were 36.5% and 5.0% respectively19. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (LR+), negative likelihood ratio (LR−), and their 95% confidence intervals were calculated for Δ-ΔPCO2 and ΔScvO2.

Variables are usually considered of good clinical tool (having good discriminative property tests) when the inferior limits of the 95% confidence interval of their AUC are more than 0.7520. For this purpose, 43 patients would be sufficient for a power of 80% and an alpha risk of 0.05. Statistical analysis was performed using STATA 14.0 (StataCorp LP, College Station, Texas, USA). p < 0.05 was considered statistically significant. All reported p values are 2-sided.

Results

We studied 49 patients whose characteristics are summarized in Table 1. Twenty-five of the 49 patients (51%) were defined as responders because thermodilution CI increased by > 10% after VE of 500-ml.

Table 1 Baseline characteristics of the study population.
Full size table

There were no significant differences in patient characteristics, SOFA score, and norepinephrine between responders and non-responders (Table 1).

Effect of volume expansion on hemodynamic variables

At baseline, all the tested hemodynamic variables were similar between the two groups (Table 2). VE significantly increased arterial pressures, CVP, and intrathoracic blood volume, and decreased hemoglobin in both groups. CI and stroke volume index increased significantly only in responders after VE, whereas HR and extravascular lung water did not change (Table 2). Arterial and venous pH were not significantly different between responders and non-responders’ groups at baseline, and did not change significantly after VE (Table 2).

Table 2 Hemodynamic and acid–base variables before and after 500 ml of volume expansion.
Full size table

Effect of volume expansion on oxygenation and CO2-derived variables

At baseline, there were no significant differences between the responders and non-responders’ groups regarding all the oxygenation and CO2-derived variable (Table 3). DO2 and ScvO2 increased significantly, and OE decreased only in responders’ group after VE. VE significantly reduced ΔPCO2 and lactate levels only in the responders’ group (Table 3). VO2 and ΔPCO2/ΔContO2 ratio remained unchanged in both groups after VE.

Table 3 Oxygenation and CO2-derived variables before and after 500 ml of volume expansion.
Full size table

We observed significant correlations between ΔScvO2 and ΔCI (r = 0.42, p = 0.003) and between Δ-ΔPCO2 and ΔCI (r =  − 0.30, p = 0.03) after VE.

In patients with an increase in VO2 ≤ 10% (n = 36) after VE, the correlation between ΔScvO2 and ΔCI was of 0.62 (p < 0.001). Also, in these patients, Δ-ΔPCO2 was significantly correlated with ΔCI (r =  − 0.47, p = 0.004).

Ability of ΔScvO2 and Δ-ΔPCO2 to define fluid responsiveness (increase in CI > 10% after VE)

The AUC for ΔScvO2 was 0.68 (95% CI: 0.53–0.83) (p = 0.02) and for Δ-ΔPCO2 was 0.76 (95% CI: 0.63–0.89) (p < 0.001) (Fig. 1). There were no significant differences between the AUCs for ΔScvO2 and Δ-ΔPCO2 (p = 0.41).

Figure 1
figure 1

Receiver operating characteristic (ROC) curves showing the ability of the changes in ΔPCO2 (Δ-ΔPCO2) (green curve), ScvO2 (ΔScvO2) (blue curve) between before and after 500 mL of volume expansion to define fluid responsiveness (increase in cardiac index > 10% after volume expansion).

Full size image

The best cutoff value (according to Youden index) for ΔScvO2 was ≥ 3.5% (sensitivity = 64% [95% CI: 42–82%], specificity = 65% [95% CI: 43–84%]), which was lower than its LSC (5%). Taking into account the repeatability (LSC), the best cutoff value was ≥ 7.4% (sensitivity = 56% [95% CI: 35–76%], specificity = 71% [95% CI: 49–87%], PPV = 63% [95% CI: 39–83%], NPV = 61% [95% CI: 41–78%], LR+ = 1.8 [95% CI: 0.9–3.7], and LR− = 0.7 [95% CI: 0.4–1.1]).

The best cutoff value (according to Youden index) for Δ-ΔPCO2 was ≤ -23.5% (sensitivity = 52% [95% CI: 31–72%], specificity = 87% [95% CI: 68–97%], which was lower than its LSC (36.5%). Taking into account the repeatability (LSC), the best cutoff value was ≤ -37.5% (sensitivity = 32% [95% CI:15–53%], specificity = 92% [95% CI: 73–99%], PPV = 79% [95% CI: 42–97%], NPV = 58% [95% CI: 42–74%], LR+ = 3.8 [95% CI: 0.9–16.3], and LR− = 0.7 [95% CI:0.6–1.0]).

In patients with an increase in VO2 ≤ 10%, the AUC for ΔScvO2 was 0.73 (95% CI: 0.57–0.90) (p = 0.006) and for Δ-ΔPCO2 was 0.83 (95% CI: 0.69–0.97) (p < 0.001) (Fig. 2). There was no significant difference between the AUCs for ΔScvO2 and Δ-ΔPCO2 (p = 0.41).

Figure 2
figure 2

Receiver operating characteristic (ROC) curves showing the ability of the changes in ΔPCO2 (Δ-ΔPCO2) (green curve), ScvO2 (ΔScvO2) (blue curve) between before and after 500 mL of volume expansion to define fluid responsiveness (increase in cardiac index > 10% after volume expansion) in the subgroup of patients with an increase in oxygen consumption ≤ 10%.

Full size image

The best cutoff value (according to Youden index) for ΔScvO2 was ≥ 8.1% (sensitivity = 65% [95% CI: 38–86%], specificity = 74% [95% CI: 49–91%]), PPV = 73% [95% CI: 47–91%], NPV = 65% [95% CI: 40–85%], LR+ = 2.5 [95% CI: 1.1–5.6], and LR− = 0.5 [95% CI: 0.2–1.0]).

The best cutoff value (according to Youden index) for Δ-ΔPCO2 was ≤ -25% (sensitivity = 59% [95% CI: 33–82%], specificity = 89% [95% CI: 65–99%], which was lower than its LSC (36.5%). Taking into account the repeatability (LSC), the best cutoff value was ≤  − 37.5% (sensitivity = 41% [95% CI: 18–67%], specificity = 100% [95% CI: 59–100%], PPV = 100% [95% CI: 62–100%], NPV = 60% [95% CI: 40–78%], LR+ = ∞, and LR− = 0.60 [95% CI: 0.4–0.9]).

Characteristics of patients with tissue hypoxia (ΔVO2 > 10% after VE)

At baseline, ΔPCO2/ΔContO2 ratio was significantly higher in patients with ΔVO2 > 10% (tissue hypoxia) induced by volume expansion compared to patients with ΔVO2 ≤ 10% (2.03 [1.73–2.27] vs. 1.39 [1.00–1.71] mmHg/mL, p = 0.03, respectively). We did not observe significant differences between patients with ΔVO2 > 10% and ΔVO2 ≤ 10% regarding baseline lactate levels (2.35 [1.02–3.97] vs. 1.70 [1.40–3.10] mmol/L, p = 0.89, respectively) and baseline ScvO2 levels (60 ± 17% vs. 59 ± 13%, p = 0.78, respectively).

The AUCs of baseline lactate and ScvO2 values to predict ΔVO2 > 10% were 0.48 (95% CI: 0.28–0.69) (p = 0.89) and 0.54 (95% CI: 0.33–0.74) (p = 0.72), respectively. The AUC of baseline ΔPCO2/ΔContO2 ratio to predict ΔVO2 > 10% after volume expansion was 0.84 (95% CI: 0.71–0.96) (p < 0.001) (Fig. 3). The best cutoff value (according to Youden index) for baseline ΔPCO2/ΔContO2 ratio was > 1.70 (sensitivity = 77% [95% CI: 46–95%], specificity = 77% [95% CI: 60–90%]), PPV = 90% [95% CI: 73–98%], NPV = 55% [95% CI :31–78%], LR+ = 3.4 [95% CI: 1.7–6.6], and LR− = 0.3 [95% CI: 0.1–0.8]).

Figure 3
figure 3

Receiver operating characteristic (ROC) curves showing the ability of baseline ΔPCO2/ΔContO2 (red curve), ScvO2 (yellow curve), and lactate (black curve) to predict an increase in oxygen consumption > 10% after 500 mL of volume expansion.

Full size image

Discussion

The main findings of our study are: (1) in the whole population Δ-ΔPCO2 induced by VE has an acceptable ability to define fluid responsiveness, but not ΔScvO2; (2) The abilities of Δ-ΔPCO2 and ΔScvO2 to define fluid responsiveness improved when we considered only patients in whom changes in VO2 were minimal (ΔVO2 ≤ 10%) after VE, i.e., patients without tissue hypoxia; 3) Baseline ΔPCO2/ΔContO2 ratio has a good ability to predict the presence of tissue hypoxia (increases in VO2 > 10% after VE).

Applying the modified Fick method to CO2, venous-to-arterial PCO2 difference reflects cardiac output. Several experimental studies have demonstrated the primary role of decreased tissue blood flow in the increased venous-to-arterial PCO2 gap8,21,22. Similarly, a mathematical model analysis has confirmed that blood flow represents the major determinant in the elevation of venous-to-arterial PCO2 gap23. A rise in mixed venous-to-arterial PCO2 gap that was directly linked to a decrease in cardiac output has been observed in different types of circulatory failure including septic shock24,25. Mecher et al. found a significant negative correlation between the changes in cardiac output and mixed venous-to-arterial PCO2 gap after fluid resuscitation in septic shock patients (r =  − 0.42, p < 0.01)24. In post-cardiac surgery sedated and mechanically ventilated patients with cardiac index < 2.5 L/min/m2, ΔCI induced by VE (500-mL bolus of crystalloid given over 30 min) was found to be significantly correlated with Δ-ΔPCO2 (r =  − 0.53, p = 0.001)13. However, the AUC for Δ-ΔPCO2 to define fluid responsiveness was not determined in that study.

The correlation between Δ-ΔPCO2 and ΔCI induced by VE was weaker in our septic patients than what was observed in post-cardiac surgery patients13. Also, the ability of Δ-ΔPCO2 to define fluid responsiveness was not robust (AUC = 0.76). These findings could be explained by several factors. First, the relationship between ΔPCO2 and CI is curvilinear9, which means that the magnitude of changes in ΔPCO2 is more pronounced at low CI than at normal or high CI. Second, the relationship between CO2 content and PCO2, which is curvilinear rather than linear, is influenced by many factors such as the degree of metabolic acidosis, hematocrit, and oxygen saturation (Haldane effect)9,26. However, we believe that this factor is unlikely to have occurred in our patients. Although base excess and hemoglobin significantly decreased in both groups (responders and non-responders) and ScvO2 increased only in the responders’ group (Tables 2, 3) after VE, it is unlikely that these clinically irrelevant changes could have affected the PCO2/CO2 content relationship. If these changes had affected the PCO2/CO2 content relationship, it would have resulted in an increase in ΔPCO2 in both groups. Third, in situations of tissue hypoxia with VO2/DO2 dependency phenomenon and anaerobic CO2 production, the rise in CI would increase VO2 and VCO2. This would attenuate the decrease in ΔPCO2 related to the increase in blood flow9,27. We believe that this factor may have contributed to the reduction in the performance of Δ-ΔPCO2 (AUC = 0.76) in defining an increase in CI > 10% induced by VE (fluid responsiveness) in the overall population. When we excluded patients with tissue hypoxia, patients with an increase in VO2 ≤ 10% (VO2/DO2 independency), we observed an improvement in the ability of Δ-ΔPCO2 to define fluid responsiveness with a very good AUC of 0.83 (Fig. 2). A decrease in Δ-ΔPCO2 of more or equal than 37.5% induced by VE allowed discrimination between responders and non-responders with a PPV of 100%. Also, the correlation between Δ-ΔPCO2 and ΔCI was higher than in the overall population.

Venous oxygen saturation is a global marker of adequacy between oxygen consumption and oxygen supply28. Therefore, changes in venous oxygen saturation reflect changes in the balance between VO2 and DO2 and indicate tissue oxygenation. Giraud et al. observed, in 30 cardiogenic shock or postoperative cardiac surgery patients, that ΔScvO2 was significantly correlated with ΔCI induced by a bolus of 500 mL of normal saline administered over 10-min (r = 0.67, p < 0.001)14. Also, ΔScvO2 had an excellent ability to define an increase in CI ≥ 15% after VE (fluid responsiveness) with an AUC of 0.90. Our findings are different from those reported by Giraud et al.14. We observed in our whole septic population a weaker correlation between ΔScvO2 and ΔCI (r = 0.42), and ΔScvO2 had a poor ability to discriminate between responders and non-responders. In the subgroup of patients without tissue hypoxia (no significant increase in VO2, or VO2/DO2 independency), even though the correlation between ΔScvO2 and ΔCI improved, the ability of ΔScvO2 to characterize fluid responsiveness was not good. The main explanation of the discrepancies between our findings and those of Giraud et al.14 is that the patient populations are different. As has been previously described, venous oxygen saturation may not be a good indicator of tissue oxygenation in the setting of sepsis, due to microcirculatory shunting and mitochondrial dysfunction that can result in oxygen extraction abnormalities29.

It has been suggested that ΔPCO2/ΔContO2 ratio, considered as a surrogate of the respiratory quotient, can be used as a marker of global tissue hypoxia in critically ill patients15,16,30. In our study, we found that baseline ΔPCO2/ΔContO2 value was significantly higher in patients with global tissue hypoxia (defined as an increase in VO2 > 10% after VE) than those without global tissue hypoxia. Also, baseline ΔPCO2/ΔContO2 ratio had a very good ability to predict the presence of VO2/DO2 dependency (global tissue hypoxia). Our results confirmed our previous findings15 and those reported by Monnet et al.16, who observed excellent predictability of baseline ΔPCO2/ΔContO2 value for tissue hypoxia (AUCs = 0.96 and 0.94, respectively). Baseline lactate and ScvO2 levels had poor ability to predict VO2/DO2 dependency (global tissue hypoxia) in our study. This finding is in line of what we observed previously in septic shock patients15.

To the best of our knowledge, our study is the first to investigate the role of Δ-ΔPCO2 and ΔScvO2 in defining fluid responsiveness in septic patients. Our findings are valuable as they can be integrated in a clinical algorithm and used by the bedside provider as part of the assessment of fluid responsiveness. These values are readily available as patients in septic shock usually have arterial and central venous catheters inserted. After measuring O2 content and CO2 partial pressures, the ΔPCO2/ΔContO2 ratio can help the provider recognize patients without tissue hypoxia. In these patients, ΔPCO2 can be measured before and after VE and be used to appreciate if the latter has resulted in a significant increase in CI and to guide further fluid resuscitation when cardiac output monitoring is not available.

Our study presents several limitations. First, it is a single-center study, so the results might not universally apply. Second, we used central venous samples instead of mixed venous to assess oxygen and CO2-derived variables, limiting its accuracy. However, we were interested in the changes in these variables induced by fluid challenge rather than their absolute values. Moreover, it has been shown that calculating the oxygen and CO2-derived variables from the central venous blood permitted the detection of global tissue hypoxia in critically ill patients15,16. Third, our patients were sedated and mechanically ventilated with stable oxygen consumption; thus, our findings might not apply to spontaneously breathing patients with varying oxygen demands. Finally, our study was not sufficiently powered for subgroup analyses; thus, our findings need to be replicated in a future study with larger sample size.

Conclusions

In sedated and mechanically ventilated septic patients with no signs of tissue hypoxia, Δ-ΔPCO2 is a reliable parameter to define fluid responsiveness and can be used in the absence of CI measurement. Baseline ΔPCO2/ΔContO2 ratio could help the physician recognize the presence of tissue hypoxia.