Venous Minus Arterial Carbon Dioxide Gradients in the Monitoring of Tissue Perfusion and Oxygenation: A Narrative Review

According to Fick’s principle, the total uptake of (or release of) a substance by tissues is the product of blood flow and the difference between the arterial and the venous concentration of the substance. Therefore, the mixed or central venous minus arterial CO2 content difference depends on cardiac output (CO). Assuming a linear relationship between CO2 content and partial pressure, central or mixed venous minus arterial PCO2 differences (Pcv-aCO2 and Pmv-aCO2) are directly related to CO. Nevertheless, this relationship is affected by alterations in the CO2Hb dissociation curve induced by metabolic acidosis, hemodilution, the Haldane effect, and changes in CO2 production (VCO2). In addition, Pcv-aCO2 and Pmv-aCO2 are not interchangeable. Despite these confounders, CO is a main determinant of Pcv-aCO2. Since in a study performed in septic shock patients, Pmv-aCO2 was correlated with changes in sublingual microcirculation but not with those in CO, it has been proposed as a monitor for microcirculation. The respiratory quotient (RQ)—RQ = VCO2/O2 consumption—sharply increases in anaerobic situations induced by exercise or critical reductions in O2 transport. This results from anaerobic VCO2 secondary to bicarbonate buffering of anaerobically generated protons. The measurement of RQ requires expired gas analysis by a metabolic cart, which is not usually available. Thus, some studies have suggested that the ratio of Pcv-aCO2 to arterial minus central venous O2 content (Pcv-aCO2/Ca-cvO2) might be a surrogate for RQ and tissue oxygenation. In this review, we analyze the physiologic determinants of Pcv-aCO2 and Pcv-aCO2/Ca-cvO2 and their potential usefulness and limitations for the monitoring of critically ill patients. We discuss compelling evidence showing that they are misleading surrogates for tissue perfusion and oxygenation, mainly because they are systemic variables that fail to track regional changes. In addition, they are strongly dependent on changes in the CO2Hb dissociation curve, regardless of changes in systemic and microvascular perfusion and oxygenation.


Introduction
The monitoring of the adequacy of tissue perfusion and oxygenation is a major task in the assessment of critically ill patients. Unfortunately, few tools are available for these goals. The clinical evaluation of skin perfusion by means of the capillary refill time is a valuable method [1]. It is a cheap and easy technique, which can be performed in different sites, such as the fingertip (pulp or nail), earlobe, thumb, forehead, and chest wall. In healthy volunteers, there is a good agreement between capillary refill time measured in the pulp fingertip and the ear lobe [2]. The measurement of capillary refill time, however, is poorly by direct tonometry is extremely cumbersome. On the other hand, the calculation of CO 2 content depends on complex formulae that frequently produce unacceptable errors. The method more commonly used was allegedly validated in comparison with manometric measurements performed by the Van Slyke method [19]. The authors found an excellent correlation between both determinations. Even though, using data provided in the manuscript, it is possible to calculate the 95% limits of agreement between calculated and measured CO 2 content. The resulting value is 4.66 mL/100 mL, which is very wide. Thus, the methods are not interchangeable, especially considering the error propagation related to the calculation of C v-a CO 2 . Accordingly, 5-10% of the calculated C v-a CO 2 values are negative, which is not physiologically possible. Improved algorithms for the calculation of CO 2 content have been developed, but they still show inaccuracies [20,21].
Taking into account these drawbacks, P v-a CO 2 is commonly used instead of C v-a CO 2 . The relationship between CO 2 content and partial pressure, however, is not straightforward and depends on several factors ( Figure 1): hemorrhagic shock does not increase the intestinal mucosal Pt-aCO2 because of the reduction in the VCO2 [18]. Another problematic issue related to the clinical usefulness of Fick's principle applied to CO2 for the monitoring of blood flow is the measurement of CO2 content. Determination by direct tonometry is extremely cumbersome. On the other hand, the calculation of CO2 content depends on complex formulae that frequently produce unacceptable errors. The method more commonly used was allegedly validated in comparison with manometric measurements performed by the Van Slyke method [19]. The authors found an excellent correlation between both determinations. Even though, using data provided in the manuscript, it is possible to calculate the 95% limits of agreement between calculated and measured CO2 content. The resulting value is 4.66 mL/100 mL, which is very wide. Thus, the methods are not interchangeable, especially considering the error propagation related to the calculation of Cv-aCO2. Accordingly, 5-10% of the calculated Cv-aCO2 values are negative, which is not physiologically possible. Improved algorithms for the calculation of CO2 content have been developed, but they still show inaccuracies [20,21].
Taking into account these drawbacks, Pv-aCO2 is commonly used instead of Cv-aCO2. The relationship between CO2 content and partial pressure, however, is not straightforward and depends on several factors ( Figure 1): (1) Position on the CO2Hb dissociation curve: Given the curvilinear characteristics of the curve, the relationship between CO2 partial pressure and content varies over the entire range of values. In the steeper portion (low PCO2), the increases in PCO2 at any CO2 content are smaller than in the flattened part (high PCO2).
(2) Haldane effect: Oxygenated Hb has a lower capacity for CO2 binding. In this way, similar CO2 content is associated with higher PCO2 at higher oxygen saturations [22]. This mechanism favors the Hb loading of CO2 produced by the tissue metabolism in the peripheral capillaries and its unloading in the lungs. Although the PCO2 only falls from 45 mmHg on the venous side to 40 mmHg on the arterial side, the CO2 content decreases by a much greater extent ( Figure 1).
(3) Effect of acidosis: Metabolic acidosis decreases the Hb ability to transport CO2 [23]. (1) Position on the CO 2 Hb dissociation curve: Given the curvilinear characteristics of the curve, the relationship between CO 2 partial pressure and content varies over the entire range of values. In the steeper portion (low PCO 2 ), the increases in PCO 2 at any CO 2 content are smaller than in the flattened part (high PCO 2 ).
(2) Haldane effect: Oxygenated Hb has a lower capacity for CO 2 binding. In this way, similar CO 2 content is associated with higher PCO 2 at higher oxygen saturations [22]. This mechanism favors the Hb loading of CO 2 produced by the tissue metabolism in the peripheral capillaries and its unloading in the lungs. Although the PCO 2 only falls from 45 mmHg on the venous side to 40 mmHg on the arterial side, the CO 2 content decreases by a much greater extent ( Figure 1).
Considering these mechanisms, P v-a CO 2 and P t-a CO 2 not only depend on blood flow and VCO 2 but also on changes in the CO 2 Hb dissociation curve ( Figure 2). Shifts in the CO 2 Hb dissociation curve can induce major changes in those differences.
(5) Temperature: Increases in temperature induce a right shift in the HbCO2 dissociation curve [25].
Considering these mechanisms, Pv-aCO2 and Pt-aCO2 not only depend on blood flow and VCO2 but also on changes in the CO2Hb dissociation curve (Figure 2). Shifts in the CO2Hb dissociation curve can induce major changes in those differences. Determinants of venous minus arterial and tissue minus arterial PCO2 differences). Venous minus arterial and tissue minus arterial PCO2 differences (ΔPCO2) are the result of interactions among CO2 production (VCO2), CO2Hb dissociation curve, and blood flow. Isolated changes in any determinant can independently modify PCO2 differences.
Another relevant concept is that CO2 gradients are determined by flow, not by DO2. Despite similar degrees of oxygen supply dependence in isolated hindlimbs, regional Pv-aCO2 increased more than twofold in ischemic hypoxia and remained unchanged in hypoxic hypoxia, in which blood flow is normal [26]. Similar findings were described in whole animal models of hypoxic and anemic hypoxia, in which not only systemic and regional Pv-aCO2 but also Pt-aCO2 failed to reflect tissue hypoxia [27][28][29]. In both situations, blood flow is preserved. Therefore, CO2 differences depend on flow, and not on tissue hypoxia.

Venous Minus Arterial Carbon Dioxide Partial Pressure in Shock States
During the reductions in CO, there are opposite changes in O2 and CO2 venous content. Low-flow states are characterized by low venous O2 saturation and high venous PCO2. In low CO states, tissue and venous hypercarbia are ubiquitous phenomena that arise as a consequence of the reduced washout of CO2. In the eighties, the occurrence of venous hypercarbia during cardiac arrest was well-documented [30][31][32]. Experimental and clinical studies also found a widened Pv-aCO2 in other low CO states, such as hemorrhagic shock [33][34][35] and cardiac failure [32]. In hemorrhagic shock, Pv-aCO2 predictably reflects changes in CO. In acute progressive bleeding, the reductions in CO induce semilogarithmic increases in Pmv-aCO2 ( Figure 3) [28]. This regression fitting was repeatedly found in several conditions [36][37][38]. Determinants of venous minus arterial and tissue minus arterial PCO 2 differences). Venous minus arterial and tissue minus arterial PCO 2 differences (∆PCO 2 ) are the result of interactions among CO 2 production (VCO 2 ), CO 2 Hb dissociation curve, and blood flow. Isolated changes in any determinant can independently modify PCO 2 differences.
Another relevant concept is that CO 2 gradients are determined by flow, not by DO 2 . Despite similar degrees of oxygen supply dependence in isolated hindlimbs, regional P v-a CO 2 increased more than twofold in ischemic hypoxia and remained unchanged in hypoxic hypoxia, in which blood flow is normal [26]. Similar findings were described in whole animal models of hypoxic and anemic hypoxia, in which not only systemic and regional P v-a CO 2 but also P t-a CO 2 failed to reflect tissue hypoxia [27][28][29]. In both situations, blood flow is preserved. Therefore, CO 2 differences depend on flow, and not on tissue hypoxia.

Venous Minus Arterial Carbon Dioxide Partial Pressure in Shock States
During the reductions in CO, there are opposite changes in O 2 and CO 2 venous content. Low-flow states are characterized by low venous O 2 saturation and high venous PCO 2 . In low CO states, tissue and venous hypercarbia are ubiquitous phenomena that arise as a consequence of the reduced washout of CO 2 . In the eighties, the occurrence of venous hypercarbia during cardiac arrest was well-documented [30][31][32]. Experimental and clinical studies also found a widened P v-a CO 2 in other low CO states, such as hemorrhagic shock [33][34][35] and cardiac failure [32]. In hemorrhagic shock, P v-a CO 2 predictably reflects changes in CO. In acute progressive bleeding, the reductions in CO induce semilogarithmic increases in P mv-a CO 2 ( Figure 3) [28]. This regression fitting was repeatedly found in several conditions [36][37][38].
In experimental endotoxemic models and in patients with septic shock, P v-a CO 2 also tracks changes in CO [37,[39][40][41][42][43]. In the different studies, the strength of the correlation between P v-a CO 2 and CO was quite variable. For example, an observational study in septic patients found a weak but significant correlation between P cv-a CO 2 and CO (R 2 = 0.07, p < 0.0001) [42]. Nevertheless, the proper surrogate for CO is P cv-a CO 2 , not P mv-a CO 2 . The same study showed a poor agreement between P cv-a CO 2 and P mv-a CO 2 (95% limits of agreement = 9 mmHg), which is similar to that reported elsewhere [44]. Therefore, the variable strength of the correlation between P v-a CO 2 and CO could be explained by either modification in the other determinants (VCO 2 and HbCO 2 dissociation curve) or the use of P cv-a CO 2 instead of P mv-a CO 2 . In spite of this, P cv-a CO 2 and P mv-a CO 2 depend on CO. This expression of Fick's principle applied to CO 2 was confirmed in systematic reviews including large numbers of critically ill and septic patients [45,46].  In experimental endotoxemic models and in patients with septic shock, Pv-aCO2 also tracks changes in CO [37,[39][40][41][42][43]. In the different studies, the strength of the correlation between Pv-aCO2 and CO was quite variable. For example, an observational study in septic patients found a weak but significant correlation between Pcv-aCO2 and CO (R 2 = 0.07, p < 0.0001) [42]. Nevertheless, the proper surrogate for CO is Pcv-aCO2, not Pmv-aCO2. The same study showed a poor agreement between Pcv-aCO2 and Pmv-aCO2 (95% limits of agreement = 9 mmHg), which is similar to that reported elsewhere [44]. Therefore, the variable strength of the correlation between Pv-aCO2 and CO could be explained by either modification in the other determinants (VCO2 and HbCO2 dissociation curve) or the use of Pcv-aCO2 instead of Pmv-aCO2. In spite of this, Pcv-aCO2 and Pmv-aCO2 depend on CO. This expression of Fick's principle applied to CO2 was confirmed in systematic reviews including large numbers of critically ill and septic patients [45,46].
Given that low values of Pcv-aCO2 were associated with an improved outcome, it has been suggested as a goal for resuscitation [41,43,[45][46][47][48][49][50]. Yet, its usefulness for this purpose has never been confirmed. On the contrary, a small, controlled study showed that resuscitation aimed to improve Pcv-aCO2 increases mortality [51].
As a relevant conclusion, Pcv-aCO2 and Pmv-aCO2 are strongly dependent on CO in physiological conditions and in shock states, including septic shock. Nevertheless, the ability of these variables to track CO is dampened by many factors: (1) Haldane effect: When venous oxygen saturation increases as the result of increased blood flow, changes in venous blood CO2 partial pressure and content may differ from each other because of the Haldane effect [52]. In patients with septic shock, dobutamine-induced changes in CO were not followed by decreases in Pmv-aCO2 because of the simultaneous increase in venous O2 saturation [44].
In hyperoxia, the Haldane effect also determines increases in Pcv-aCO2 [53], even in the absence of changes in systemic and microvascular hemodynamics [54].
(2) Metabolic acidosis: The right shift in the HbCO2 dissociation curve [23] produces greater increases in PCO2 on the venous than on the arterial side. Therefore, metabolic acidosis can significantly increase Pv-aCO2 regardless of any change in blood flow [29,44,55]. Given that low values of P cv-a CO 2 were associated with an improved outcome, it has been suggested as a goal for resuscitation [41,43,[45][46][47][48][49][50]]. Yet, its usefulness for this purpose has never been confirmed. On the contrary, a small, controlled study showed that resuscitation aimed to improve P cv-a CO 2 increases mortality [51].
As a relevant conclusion, P cv-a CO 2 and P mv-a CO 2 are strongly dependent on CO in physiological conditions and in shock states, including septic shock. Nevertheless, the ability of these variables to track CO is dampened by many factors: (1) Haldane effect: When venous oxygen saturation increases as the result of increased blood flow, changes in venous blood CO 2 partial pressure and content may differ from each other because of the Haldane effect [52]. In patients with septic shock, dobutamine-induced changes in CO were not followed by decreases in P mv-a CO 2 because of the simultaneous increase in venous O 2 saturation [44].
In hyperoxia, the Haldane effect also determines increases in P cv-a CO 2 [53], even in the absence of changes in systemic and microvascular hemodynamics [54].
(2) Metabolic acidosis: The right shift in the HbCO 2 dissociation curve [23] produces greater increases in PCO 2 on the venous than on the arterial side. Therefore, metabolic acidosis can significantly increase P v-a CO 2 regardless of any change in blood flow [29,44,55].
(3) Hemodilution: Anemia also affects the ability to transport CO 2 . As repeatedly shown, hemodilution is associated with opposite changes in C v-a CO 2 and P v-a CO 2 : C v-a CO 2 decreases and P v-a CO 2 increases ( Figure 4) [28,29].
(4) Acute changes in ventilation: P mv-a CO 2 increases with hyperventilation and decreases with hypoventilation [52,56,57]. Underlying mechanisms might be the reduction in blood flow and the increase in VCO 2 driven by systemic alkalosis [58].
(5) Changes in temperature: Changes in body temperature induce parallel modifications in oxidative metabolism and VCO 2 [18].
(6) Use of central instead of mixed venous samples: There are wide 95% limits of agreement between calculations of P v-a CO 2 using central or mixed venous blood [42,44]. Thus, P cv-a CO 2 might not reflect CO as well as P mv-a CO 2 .
(7) The variability of the measurements: Given the variability of the measurements in successive determinations of the P v-a CO 2 gap, it is recommended to consider only variations of at least ±2 mmHg as real changes [59]. (3) Hemodilution: Anemia also affects the ability to transport CO2. As repeatedly shown, hemodilution is associated with opposite changes in Cv-aCO2 and Pv-aCO2: Cv-aCO2 decreases and Pv-aCO2 increases ( Figure 4) [28,29]. Relationship of systemic oxygen transport to mixed venous minus arterial PCO2 difference (A), mixed venous minus arterial CO2 content difference (B), and arterial minus mixed venous oxygen content difference (C) in sheep that underwent progressive bleeding or hemodilution. In hemorrhage, the three variables increased. In hemodilution, there were opposite changes in mixed venous minus arterial CO2 partial pressure and content difference (the former increased, and the latter decreased), whereas arterial minus mixed venous oxygen content difference decreased. Reproduced from Ref. [29] with permission.
(4) Acute changes in ventilation: Pmv-aCO2 increases with hyperventilation and decreases with hypoventilation [52,56,57]. Underlying mechanisms might be the reduction in blood flow and the increase in VCO2 driven by systemic alkalosis [58].
(5) Changes in temperature: Changes in body temperature induce parallel modifications in oxidative metabolism and VCO2 [18].
(6) Use of central instead of mixed venous samples: There are wide 95% limits of agreement between calculations of Pv-aCO2 using central or mixed venous blood [42,44]. Thus, Pcv-aCO2 might not reflect CO as well as Pmv-aCO2.
(7) The variability of the measurements: Given the variability of the measurements in successive determinations of the Pv-aCO2 gap, it is recommended to consider only variations of at least ± 2 mmHg as real changes [59].

Venous Minus Arterial Carbon Dioxide Partial Pressure as a Monitor of Microcirculatory Perfusion in Septic Shock
Septic shock is a condition in which the coherence between systemic hemodynamics and microcirculation can be lost. A systemic hyperdynamic state can coexist with microvascular hypoperfusion in some territories. Tissue hypoperfusion could be identified by means of Pt-aCO2. Accordingly, experimental and clinical studies showed that sublingual, intestinal mucosal, and cutaneous Pt-aCO2 correlate with the respective microcirculatory flow [60][61][62]. In contrast, the systemic Pv-aCO2 depends on CO, while the regional Pv-aCO2 of different organs is determined by the corresponding blood flow of each organ. In conditions characterized by the dissociation between systemic cardiovascular variables and microcirculation, systemic Pv-aCO2 is also dissociated from Pt-aCO2 and microcirculation. Thus, systemic variables, such as Pmv-aCO2 and Pcv-aCO2 could fail to reflect tissue hypoperfusion. Nevertheless, many reviews recommended the use of Pcv-aCO2 for the monitoring of microcirculation in critically ill patients, even in situations of normal or high CO [15,49,[63][64][65][66][67]. This recommendation is only based on the results of an observational study, which assessed the relationship of Pmv-aCO2 to systemic hemodynamics and sublingual microcirculation [66]. Seventy-five patients with septic shock were evaluated at basal conditions and 6 h later. The study showed that changes in Pmv-aCO2 correlated with changes in the proportion of perfused microvessels, but there was no such correlation between Pmv-aCO2 and CO. The main conclusion of the study was that Pmv-aCO2 could reflect Relationship of systemic oxygen transport to mixed venous minus arterial PCO 2 difference (A), mixed venous minus arterial CO 2 content difference (B), and arterial minus mixed venous oxygen content difference (C) in sheep that underwent progressive bleeding or hemodilution. In hemorrhage, the three variables increased. In hemodilution, there were opposite changes in mixed venous minus arterial CO 2 partial pressure and content difference (the former increased, and the latter decreased), whereas arterial minus mixed venous oxygen content difference decreased. Reproduced from Ref. [29] with permission.

Venous Minus Arterial Carbon Dioxide Partial Pressure as a Monitor of Microcirculatory Perfusion in Septic Shock
Septic shock is a condition in which the coherence between systemic hemodynamics and microcirculation can be lost. A systemic hyperdynamic state can coexist with microvascular hypoperfusion in some territories. Tissue hypoperfusion could be identified by means of P t-a CO 2 . Accordingly, experimental and clinical studies showed that sublingual, intestinal mucosal, and cutaneous P t-a CO 2 correlate with the respective microcirculatory flow [60][61][62]. In contrast, the systemic P v-a CO 2 depends on CO, while the regional P v-a CO 2 of different organs is determined by the corresponding blood flow of each organ. In conditions characterized by the dissociation between systemic cardiovascular variables and microcirculation, systemic P v-a CO 2 is also dissociated from P t-a CO 2 and microcirculation. Thus, systemic variables, such as P mv-a CO 2 and P cv-a CO 2 could fail to reflect tissue hypoperfusion. Nevertheless, many reviews recommended the use of P cv-a CO 2 for the monitoring of microcirculation in critically ill patients, even in situations of normal or high CO [15,49,[63][64][65][66][67]. This recommendation is only based on the results of an observational study, which assessed the relationship of P mv-a CO 2 to systemic hemodynamics and sublingual microcirculation [66]. Seventy-five patients with septic shock were evaluated at basal conditions and 6 h later. The study showed that changes in P mv-a CO 2 correlated with changes in the proportion of perfused microvessels, but there was no such correlation between P mv-a CO 2 and CO. The main conclusion of the study was that P mv-a CO 2 could reflect microvascular flow and not systemic hemodynamic variables. Considering that this suggestion challenges Fick's principle, the lack of correlation between P mv-a CO 2 and CO should have been explained by changes in the many other determinants of P mv-a CO 2 , mainly those that modify the dissociation of CO 2 from Hb. The authors stated that corrections for the Haldane effect were done, but this point was not clearly addressed in the manuscript, especially because O 2 saturations were calculated instead of being directly measured by a co-oximeter.
Another study, performed in patients with cardiogenic shock on venoarterial extracorporeal membrane oxygenation, found that P v-a CO 2 was higher in nonsurvivors than in survivors (7.4 mm Hg [5.7-10.1] vs. 5.9 mm Hg [3.8-9.2], p < 0.01) [68]. Since the flow rate was similar in both groups, the authors concluded that a high P v-a CO 2 might reveal the presence of a microcirculatory dysfunction. Regardless of the subtle difference in P v-a CO 2 between groups, the study showed a correlation between P v-a CO 2 and flow rate. Moreover, venous oxygen saturation and lactate were higher and hemoglobin was lower in nonsurvivors than in survivors. In the absence of direct microvascular assessment, differences in P v-a CO 2 could be completely explained by these findings. Consequently, any reference to microcirculatory dysfunction may be reasonable but also speculative.
Contrary to the intriguing findings and interpretations of those studies [66,68], a large body of evidence shows that P v-a CO 2 and CO are correlated in septic shock [37,[39][40][41][42][43]45,46]. Moreover, several studies showed that systemic and regional P v-a CO 2 fail to reflect microvascular perfusion because they are dependent on systemic or regional flow, and not on microvascular perfusion. In an experimental model of septic shock, the administration of endotoxin initially induced a hypodynamic state with reductions in CO, superior mesenteric artery blood flow, and mucosal microcirculatory perfusion. This condition was indicated by the widening of systemic, regional, and tissue PCO 2 gradients [60]. Fluid resuscitation increased CO and superior mesenteric artery blood flow but failed to improve villi microcirculation. Accordingly, systemic and intestinal P v-a CO 2 normalized. In contrast, mucosal P t-a CO 2 remained elevated as an expression of the persistent villi hypoperfusion [60] (Figure 5). In patients with septic shock, sublingual microcirculation was altered and red blood cell velocity was low regardless of the systemic hemodynamic pattern [69]. P mv-a CO 2 , however, was lower in patients with hyperdynamic shock (cardiac index ≥ 4.0 L/min/m 2 ) than in patients with normal CO (7 ± 2 vs. 5 ± 3 mm Hg, p < 0.05) ( Figure 6). Another study, performed in patients with septic shock, found that skin flow was correlated with the cutaneous P t-a CO 2 and was a strong predictor of outcome. As an expression of the lack of coherence between systemic hemodynamics and microcirculation, skin perfusion did not correlate with CO, and neither CO nor P mv-a CO 2 was a predictor of outcome [62]. Unrelated to P v-a CO 2 , P t-a CO 2 does track changes in microvascular perfusion [60][61][62].
between groups, the study showed a correlation between Pv-aCO2 and flow rate. Moreover, venous oxygen saturation and lactate were higher and hemoglobin was lower in nonsurvivors than in survivors. In the absence of direct microvascular assessment, differences in Pv-aCO2 could be completely explained by these findings. Consequently, any reference to microcirculatory dysfunction may be reasonable but also speculative.
Contrary to the intriguing findings and interpretations of those studies [66,68], a large body of evidence shows that Pv-aCO2 and CO are correlated in septic shock [37,[39][40][41][42][43]45,46]. Moreover, several studies showed that systemic and regional Pv-aCO2 fail to reflect microvascular perfusion because they are dependent on systemic or regional flow, and not on microvascular perfusion. In an experimental model of septic shock, the administration of endotoxin initially induced a hypodynamic state with reductions in CO, superior mesenteric artery blood flow, and mucosal microcirculatory perfusion. This condition was indicated by the widening of systemic, regional, and tissue PCO2 gradients [60]. Fluid resuscitation increased CO and superior mesenteric artery blood flow but failed to improve villi microcirculation. Accordingly, systemic and intestinal Pv-aCO2 normalized. In contrast, mucosal Pt-aCO2 remained elevated as an expression of the persistent villi hypoperfusion [60] (Figure 5). In patients with septic shock, sublingual microcirculation was altered and red blood cell velocity was low regardless of the systemic hemodynamic pattern [69]. Pmv-aCO2, however, was lower in patients with hyperdynamic shock (cardiac index ≥ 4.0 L/min/m 2 ) than in patients with normal CO (7 ± 2 vs. 5 ± 3 mm Hg, p < 0.05) ( Figure 6). Another study, performed in patients with septic shock, found that skin flow was correlated with the cutaneous Pt-aCO2 and was a strong predictor of outcome. As an expression of the lack of coherence between systemic hemodynamics and microcirculation, skin perfusion did not correlate with CO, and neither CO nor Pmv-aCO2 was a predictor of outcome [62]. Unrelated to Pv-aCO2, Pt-aCO2 does track changes in microvascular perfusion [60][61][62].  Figure 5. Failure of venous minus arterial PCO 2 difference (P mv-a CO 2 ) to reflect tissue perfusion in an experimental model of endotoxemic shock and fluid resuscitation. In experimental septic shock, the administration of endotoxin initially induced a hypodynamic state with reductions in cardiac output, superior mesenteric artery blood flow, and mucosal microcirculatory perfusion. This condition was indicated by the widening of systemic, regional, and tissue PCO 2 gradients. Fluid resuscitation increased cardiac output and superior mesenteric artery blood flow but failed to improve villi microcirculation. Accordingly, systemic and intestinal venous minus arterial PCO 2 difference (P v-a CO 2 ) normalized. In contrast, mucosal tissue minus arterial PCO 2 (P t-a CO 2 ) remained elevated as an expression of the persistent villi hypoperfusion (From data of Ref. [60]). the administration of endotoxin initially induced a hypodynamic state with reductions in cardiac output, superior mesenteric artery blood flow, and mucosal microcirculatory perfusion. This condition was indicated by the widening of systemic, regional, and tissue PCO2 gradients. Fluid resuscitation increased cardiac output and superior mesenteric artery blood flow but failed to improve villi microcirculation. Accordingly, systemic and intestinal venous minus arterial PCO2 difference (Pv-aCO2) normalized. In contrast, mucosal tissue minus arterial PCO2 (Pt-aCO2) remained elevated as an expression of the persistent villi hypoperfusion (From data of Ref. [60]). . The histograms of patients with normo-and hyperdynamic septic shock were similar and shifted to the left (lower velocities). Nevertheless, the mixed venous minus arterial PCO2 difference was higher in normo-than in hyperdynamic patients (7 ± 2 vs. 5 ± 3 mm Hg, p < 0.05). Reprinted from Ref. [69] with permission of the American Thoracic Society. Copyright © 2023 American Thoracic Society. All rights reserved.

Physiological Background
Under aerobic conditions, progressive workloads of exercise are associated with equivalent rises in VCO2 and VO2 as a reflection of the increasing oxidative metabolism. Therefore, the slope of the relationship-the RQ-persists initially unchanged. When the exercise becomes anaerobic, however, the increases in VCO2 surpass those from VO2, and the RQ abruptly increases. This phenomenon concurs with the occurrence of hyperlactatemia and is known as the anaerobic threshold [70]. In the other extreme of physiology, during oxygen supply dependence, the RQ sharply rises because the decreases in VO2 are Nevertheless, the mixed venous minus arterial PCO 2 difference was higher in normo-than in hyperdynamic patients (7 ± 2 vs. 5 ± 3 mm Hg, p < 0.05). Reprinted from Ref. [69] with permission of the American Thoracic Society. Copyright © 2023 American Thoracic Society. All rights reserved.

Physiological Background
Under aerobic conditions, progressive workloads of exercise are associated with equivalent rises in VCO 2 and VO 2 as a reflection of the increasing oxidative metabolism. Therefore, the slope of the relationship-the RQ-persists initially unchanged. When the exercise becomes anaerobic, however, the increases in VCO 2 surpass those from VO 2, and the RQ abruptly increases. This phenomenon concurs with the occurrence of hyperlactatemia and is known as the anaerobic threshold [70]. In the other extreme of physiology, during oxygen supply dependence, the RQ sharply rises because the decreases in VO 2 are higher than the falls in VCO 2 [11][12][13][14]. VO 2 and VCO 2 fall as an expression of the reduction in oxidative metabolism. The lower decrease in VCO 2 is explained by the appearance of anaerobic VCO 2 . In both situations, the anaerobic exercise and the critical reductions in O 2 delivery, the anaerobic VCO 2 results from the buffering by bicarbonate of anaerobically generated protons. Consequently, the increase in RQ highlights the ongoing global anaerobic metabolism. Regional RQ, calculated as C v-a CO 2 /C a-v O 2 , has also been used to determine the presence of tissue hypoxia [28,71]. In a landmark study in pigs with endotoxemic shock, the use of epinephrine-compared to norepinephrine-was associated with lower blood flow and a higher P v-a CO 2 , lactate-to-pyruvate ratio, and gastric C v-a CO 2 /C a-v O2 [71].
Of note, the evaluation of RQ and CO 2 contents is further complicated by the dynamics of CO 2 stores and the time required to reach an equilibrium after hemodynamic, ventilatory, or metabolic changes [72]. Despite the lack of complete steady-state conditions, changes in expired gases quickly provide an alert about hemodynamic and metabolic changes [11][12][13][14]70].
Even though the determination of RQ is an attractive method for the identification of global tissue hypoxia, the metabolic carts needed for its measurement are not usually available in intensive care units. Additionally, measurements of RQ are not reliable if a high inspired oxygen fraction is used [73]. For these reasons, a simplification of Fick's equation adapted to CO 2 , the P v-a CO 2 /C a-v O 2 , was proposed as a substitute for RQ [70]. Thus, high values of P v-a CO 2 /C a-v O 2 with a cutoff of 1.4 have been associated with hyperlactatemia and high mortality [74]. Furthermore, P cv-a CO 2 /C a-cv O 2 has been repeatedly included as part of algorithms for the assessment of tissue oxygenation [15,65,75,76]. Nevertheless, the evidence for these recommendations is quite limited and of low quality.
The utilization of P cv-a CO 2 /C a-cv O 2 as a surrogate for RQ and tissue oxygenation depends on the following statements. First, RQ is the ratio between VCO 2 and VO 2 : Considering Fick's equation, the previous equation can be reformulated as: Next, a similarity between mixed and central samples is taken: Then, the common factor (CO) is simplified in numerator and denominator: RQ = C cv-a CO 2 /C a-cv O 2 (4) Finally, C cv-a CO 2 is replaced by P cv-a CO 2 , assuming that CCO 2 and PCO 2 are linearly correlated over the physiological range of CO 2 content: RQ = P cv-a CO 2 /C a-cv O 2 (5) Unfortunately, some of these expectations are problematic. In the following paragraphs, these questions will be discussed.

Limitations of P cv-a CO 2 /C a-cv O 2 as a Surrogate of RQ
(1) The use of P cv-a CO 2 instead of C cv-a CO 2 in the calculation of the ratio: The investigators that proposed the utilization of P cv-a CO 2 /C a-cv O 2 as a surrogate of RQ stated that given the almost linear relationship between CO 2 content and partial pressure over the physiological range, P cv-a CO 2 is an estimate of C cv-a CO 2 in clinical practice [76]. As extensively discussed in the previous section, this asseveration is unsupported. Alterations in the CO 2 Hb dissociation curve, such as those induced by acidosis, hemodilution, and the Haldane effect, can substantially change the P cv-a CO 2 /C a-cv O 2 , regardless of the absence of alterations in RQ and tissue oxygenation. In septic patients, hyperoxia increases P cv-a CO 2 /C a-cv O 2 from 2.63 ± 1.00 to 4.34 ± 3.37 (p < 0.03) despite the lack of changes in systemic hemodynamics and sublingual microcirculation [54]. An experimental study focused on the drawbacks of P cv-a CO 2 /C a-cv O 2 as a surrogate for RQ [29]. P mv-a CO 2 /C a-mv O 2 , RQ, and their determinants were assessed during decreases in DO 2 produced by stepwise bleeding or hemodilution. P mv-a CO 2 /C a-mv O 2 and RQ were poorly correlated. Furthermore, in hemodilution, P mv-a CO 2 /C a-mv O 2 increased even before the beginning of the oxygen supply dependence and the rise in RQ ( Figure 5). This result was explained by the opposing effects of the decrease in Hb concentration on P mv-a CO 2 and C a-mv O 2 . The former increased because of the reduced ability to carry CO 2 in anemia while the latter decreased as occurs when the reduction in DO 2 depends on the fall in arterial oxygen content (Figure 7). Additionally, in the last stage of DO 2 reduction and despite comparable levels of anaerobic metabolism and increases in RQ, P mv-a CO 2 /C a-mv O 2 markedly increased in hemodilution, compared to hemorrhage, because of the abovementioned reasons. Finally, Hb, metabolic acidosis, the Haldane effect, the position in a flattened portion of the CO 2 dissociation curve, and RQ were found to be independent predictors of P mv-a CO 2 /C a-mv O 2 in a multiple linear regression model. Although P cv-a CO 2 /C a-cv O 2 was dependent on RQ, this was its weakest determinant [29]. Similar results were obtained during hypoxic hypoxia in a model of isolated hindlimb [77]. In this study, during progressive tissue hypoxia induced by hypoxemia or ischemia, P va CO 2 /C av O 2 was disproportionally higher in hypoxic than in ischemic hypoxia (almost three times in the last stage) despite similar degrees of oxygen supply dependence. Moreover, P va CO 2 /C av O 2 was higher in hypoxic than in ischemic hypoxia even before the beginning of the anaerobic metabolism.
the drawbacks of Pcv-aCO2/Ca-cvO2 as a surrogate for RQ [29]. Pmv-aCO2/Ca-mvO2, RQ, and their determinants were assessed during decreases in DO2 produced by stepwise bleeding or hemodilution. Pmv-aCO2/Ca-mvO2 and RQ were poorly correlated. Furthermore, in hemodilution, Pmv-aCO2/Ca-mvO2 increased even before the beginning of the oxygen supply dependence and the rise in RQ ( Figure 5). This result was explained by the opposing effects of the decrease in Hb concentration on Pmv-aCO2 and Ca-mvO2. The former increased because of the reduced ability to carry CO2 in anemia while the latter decreased as occurs when the reduction in DO2 depends on the fall in arterial oxygen content (Figure 7). Additionally, in the last stage of DO2 reduction and despite comparable levels of anaerobic metabolism and increases in RQ, Pmv-aCO2/Ca-mvO2 markedly increased in hemodilution, compared to hemorrhage, because of the abovementioned reasons. Finally, Hb, metabolic acidosis, the Haldane effect, the position in a flattened portion of the CO2 dissociation curve, and RQ were found to be independent predictors of Pmv-aCO2/Ca-mvO2 in a multiple linear regression model. Although Pcv-aCO2/Ca-cvO2 was dependent on RQ, this was its weakest determinant [29]. Similar results were obtained during hypoxic hypoxia in a model of isolated hindlimb [77]. In this study, during progressive tissue hypoxia induced by hypoxemia or ischemia, PvaCO2/CavO2 was disproportionally higher in hypoxic than in ischemic hypoxia (almost three times in the last stage) despite similar degrees of oxygen supply dependence. Moreover, PvaCO2/CavO2 was higher in hypoxic than in ischemic hypoxia even before the beginning of the anaerobic metabolism. , and the ratio of mixed venous minus arterial PCO2 difference to arterial minus mixed venous oxygen content difference (Pmv-aCO2/Ca-mvO2) (C) in sheep that underwent progressive bleeding or hemodilution. There were similar degrees of oxygen supply dependence and increases in the respiratory quotient in both groups. In hemodilution, however, the elevation in Pmv-aCO2/Ca-mvO2 was disproportionately higher than in hemorrhage and developed even before the development of anaerobic metabolism. Reproduced from Ref. [29] with permission.
Pcv-aCO2/Ca-cvO2 has been suggested as a tool to identify the aerobic or anaerobic origin of lactate [75,78]. As previously discussed, lactic acidosis can increase Pcv-aCO2/Ca-cvO2 because of its effects on the binding of CO2 to Hb, regardless of the aerobic or anaerobic production of lactate. In an experimental model of hemorrhagic shock, blood retransfusion normalized VO2 and RQ, but Pmv-aCO2/Ca-mvO2 remained high as a probable consequence of persistent hyperlactatemia [79]. In view of that, Pv-aCO2/Ca-vO2 could be considered a misleading tool to establish the meaning of hyperlactatemia. Similar demonstrations are required in other settings such as septic shock before generalizing this concept.
(2) The poor agreement between central and mixed venous samples: Central and mixed venous blood samples are not interchangeable for the different calculations. Although a small study advocated that mixed venous and central O2 saturation have similar behavior [80], a multicenter study demonstrated that both variables have poor agreement and that the direction of their changes over time can be different [81]. The problem is even worse for CO2-derived variables. In a clinical study, the 95% limits of agreement between Pcv-aCO2/Ca-cvO2 and Pmv-aCO2/Ca-mvO2 were 1.48, which is clinically unacceptable [44]. , and the ratio of mixed venous minus arterial PCO 2 difference to arterial minus mixed venous oxygen content difference (P mv-a CO 2 /C a-mv O 2 ) (C) in sheep that underwent progressive bleeding or hemodilution. There were similar degrees of oxygen supply dependence and increases in the respiratory quotient in both groups. In hemodilution, however, the elevation in P mv-a CO 2 /C a-mv O 2 was disproportionately higher than in hemorrhage and developed even before the development of anaerobic metabolism. Reproduced from Ref. [29] with permission. P cv-a CO 2 /C a-cv O 2 has been suggested as a tool to identify the aerobic or anaerobic origin of lactate [75,78]. As previously discussed, lactic acidosis can increase P cv-a CO 2 /C a-cv O 2 because of its effects on the binding of CO 2 to Hb, regardless of the aerobic or anaerobic production of lactate. In an experimental model of hemorrhagic shock, blood retransfusion normalized VO 2 and RQ, but P mv-a CO 2 /C a-mv O 2 remained high as a probable consequence of persistent hyperlactatemia [79]. In view of that, P v-a CO 2 /C a-v O 2 could be considered a misleading tool to establish the meaning of hyperlactatemia. Similar demonstrations are required in other settings such as septic shock before generalizing this concept.
(2) The poor agreement between central and mixed venous samples: Central and mixed venous blood samples are not interchangeable for the different calculations. Although a small study advocated that mixed venous and central O 2 saturation have similar behavior [80], a multicenter study demonstrated that both variables have poor agreement and that the direction of their changes over time can be different [81]. The problem is even worse for CO 2 -derived variables. In a clinical study, the 95% limits of agreement between P cv-a CO 2 /C a-cv O 2 and P mv-a CO 2 /C a-mv O 2 were 1.48, which is clinically unacceptable [44].
(3) The use of a defined cutoff of P cv-a CO 2 /C a-cv O 2 for the identification of the anaerobic threshold: Depending on the metabolic substrate used for oxidative metabolism, the normal RQ ranges from 0.67 to 1.30 [82]. Carbohydrate-based diet and overfeeding increase RQ while fat diet and fasting decrease RQ. In this way, the start of anaerobic metabolism is indicated by abrupt increases in RQ, not by a particular value [11][12][13][14]. The same consideration is valid for the P cv-a CO 2 /C a-cv O 2 .
(4) The use of calculated O 2 saturation for P cv-a CO 2 /C a-cv O 2 : In some studies, the computation of P cv-a CO 2 /C a-cv O 2 was performed by the use of O 2 saturation calculated from blood gases and oxyhemoglobin dissociation curve instead of measurements by co-oximetry [66,83,84]. This is a severe methodological mistake because calculated O 2 saturation is not a reliable estimate of measured values. In addition, the error of measurement is additionally propagated in the calculation of P cv-a CO 2 /C a-cv O 2 . Moreover, paired measurements of P cv-a CO 2 /C a-cv O 2 in the same analyzer are poorly reproducible with 95% limits of agreement of 1.22 [59].

The Physiological Feasibility of Increased P cv-a CO 2 /C a-cv O 2 as a Reflection of Tissue Hypoxia in Critically Ill Patients
In experiments on oxygen supply dependence, the raise in RQ is a sudden phenomenon leading to rapid death. In stepwise hemodilution, RQ rises only when Hb decreases to 1.2 g%. Similarly, in progressive hemorrhage, RQ increases when mean arterial pressure is lower than 30 mm Hg [10]. These are extreme and obvious conditions that can be easily diagnosed. High values of P cv-a CO 2 /C a-cv O 2 in adequately resuscitated patients rarely express global anaerobic metabolism. In contrast, they almost certainly result from the occurrence of factors that alter the of CO 2 Hb dissociation curve, as shown in experimental models [29] and in high-risk noncardiac surgery [85]. In both circumstances, RQ and P v-a CO 2 /C a-v O 2 showed a different behavior. In critically ill patients, a direct comparison between P cv-a CO 2 /C a-cv O 2 and RQ has not yet been performed. Therefore, values of P cv-a CO 2 /C a-cv O 2 should be cautiously interpreted in stable patients.

The Clinical Usefulness of P cv-a CO 2 /C a-cv O 2
Despite the fact that P cv-a CO 2 /C a-cv O 2 might not track the true value of RQ, it might still be useful to reflect the severity and predict the outcome of critical illness. Since it is partially determined by Hb and base excess, anemia, and metabolic acidosis can result in high P cv-a CO 2 /C a-cv O 2 by themselves and highlight the presence of a severe condition or be predictors of mortality [86,87]. Thus, anemia and metabolic acidosis might be responsible for the predictive ability of P cv-a CO 2 /C a-cv O 2 .
The ability of P cv-a CO 2 /C a-cv O 2 as a predictor of outcomes in critically ill patients has been extensively reviewed elsewhere [88]. More than twenty years ago, a retrospective study performed in 89 patients monitored with a Swan-Ganz catheter found that a value of P mv-a CO 2 /C a-mv O 2 higher than 1.4 was a predictor of hyperlactatemia and mortality [74]. Yet, P mv-a CO 2 /C a-mv O 2 values were similar in nonsurvivors and survivors (1.7 ± 1.0 vs. 1.3 ± 0.5). In contrast, lactate showed a better prognostic ability than P mv-a CO 2 /C a-mv O 2 and was higher in nonsurvivors (5.4 ± 6.1 vs. 2.0 ± 1.5 mmol/L). Despite the fact that P mv-a CO 2 /C a-mv O 2 and lactate were different over time in survivors and nonsurvivors, only C mv-a CO 2 /C a-mv O 2 and lactate, but not P mv-a CO 2 /C a-mv O 2 , were predictors of outcome in 135 patients with septic shock [83]. In another study, P cv-a CO 2 /C a-cv O 2 and lactate were lower in survivors than in nonsurvivors, but lactate was a better predictor of mortality (AUROC curves of 0.73 and 0.81, respectively) [89]. The combination of P cv-a CO 2 /C a-cv O 2 and lactate was a better predictor of mortality and organ failures than each individual variable in a retrospective study that recruited 144 patients with septic shock [84]. Additionally, in 35 patients with septic shock, P cv-a CO 2 /C a-cv O 2 was a strong predictor of lactate behavior, and both variables were associated with mortality [90]. Recent studies also found a relationship of P cv-a CO 2 /C a-cv O 2 to mortality [91][92][93].
In contrast, other studies failed to find an association between P cv-a CO 2 /C a-cv O 2 and lactate or outcome. In a large multicenter cohort study that included 363 patients with septic shock, P cv-a CO 2 /C a-cv O 2 could not differentiate patients with hyperlactatemia or poor lactate clearance from patients with normal lactate levels or adequate lactate clearance [94]. Another observational study in 23 septic patients showed that P cv-a CO 2 /C a-cv O 2 and P mv-a CO 2 /C a-mv O 2 were similar in survivors and nonsurvivors [44]. In high-risk surgical patients, RQ was a predictor of postoperative complications whereas P cv-a CO 2 /C a-cv O 2 showed no prognostic ability [85].
A recent systematic review and meta-analysis found that P cv-a CO 2 /C a-cv O 2 is associated with outcome [85]. Although the study showed little or no difference in the ability of P cv-a CO 2 /C a-cv O 2 and lactate to predict mortality, there was a trend favoring lactate. Nevertheless, the conclusions were limited by the considerable heterogeneity among the studies. After the publication of this meta-analysis, a large prospective observational study including 456 patients with septic shock compared the prognostic ability of lactate, P cv-a CO 2 , and P cv-a CO 2 /C a-cv O 2 [95]. Lactate at 6 h had the best predictive ability (AU-ROC of 0.902, 0.791, and 0.793, respectively). The combination of lactate and P cv-a CO 2 only resulted in trivial increases in the predictive value (AUROC = 0.930). In another recently published study in 98 patients with septic shock, P cv-a CO 2 /C a-cv O 2 at 24 h, but not at 8 h, was higher in nonsurvivors than in survivors and was a predictor of lactate clearance [96]. In contrast, lactate clearance was associated with outcomes at 8 h and 24 h.
Even though the relationship between P cv-a CO 2 /C a-cv O 2 and outcome is conflictive, high values of P cv-a CO 2 /C a-cv O 2 have some prognostic implications. The ability to predict mortality, however, is not superior to that of lactate. There are also controversial results about the relationship between P cv-a CO 2 /C a-cv O 2 and lactate. P cv-a CO 2 /C a-cv O 2 has also been used as a predictor of the dependence of VO 2 on DO 2 [43,97,98]. The oxygen supply dependence might indicate the occurrence of alterations in oxygen extraction and an oxygen debt, but its actual meaning is debatable [99]. Considering that VO 2 and DO 2 are usually computed from a common variable (CO), and the magnitude of change of the calculated variables is usually small, there is a considerable risk of mathematical coupling of data. Thus, oxygen supply dependence might not be an actual fact but an artifact. Moreover, those studies have a gross methodological drawback because VO 2 was calculated using central venous instead of mixed venous samples. In other studies, however, P cv-a CO 2 /C a-cv O 2 did not predict the increase in VO 2 in response to a fluid challenge [100,101]. Therefore, the evidence regarding this issue is inconclusive.
The usefulness of P cv-a CO 2 /C a-cv O 2 as a goal of resuscitation has only been assessed in two studies [47,102]. In a controlled trial, 228 septic patients were randomized to either P cv-a CO 2 /C a-cv O 2 or central venous oxygen saturation-targeted resuscitation. Mortality, organ failures, length of stay, and other secondary outcomes were similar in both groups [102]. In another small, controlled study, P cv-a CO 2 /C a-cv O 2 was not better than lactate as a goal for the resuscitation of septic patients [47].

Future Directions
The lack of correlation between P v-a CO 2 and microvascular perfusion in states of normal/high CO needs to be additionally confirmed. New studies should comprehensively assess the microcirculation and the multiple determinants of P v-a CO 2 , including changes in hemoglobin levels, acid-base status, the Haldane effect, temperature, and ventilation. Clinical research using metabolic cards, in critically ill patients, should also confirm that P cv-a CO 2 /C a-cv O 2 is poorly correlated with RQ. Furthermore, the clinical usefulness of RQ in the monitoring of critically ill patients has never been tested.

Conclusions
P v-a CO 2 and P t-a CO 2 are mainly determined by blood flow. According to Fick's principle, P mv-a CO 2 and P cv-a CO 2 are correlated with CO in physiological conditions and in critically ill patients, even in those with septic shock. Nevertheless, the relationship between CO and P v-a CO 2 is not straightforward because of the changes in the CO 2 dissociation curve and in the metabolic VCO 2 . While there is a widespread belief that P cv-a CO 2 reflects microvascular tissue perfusion, this point of view is only based on the controversial results of one observational study. The concept is mistaken because it overlooks basic physiological foundations, as well as a large body of experimental and clinical evidence. If systemic flow seems adequate, increases in P mv-a CO 2 or P cv-a CO 2 firstly indicate the presence of factors that increase the dissociation of CO 2 from Hb, such as anemia, metabolic acidosis, and the Haldane effect. In contrast, P mv-a CO 2 and P cv-a CO 2 are indicators of tissue perfusion in lowflow states. Unlike P v-a CO 2 , P t-a CO 2 does reflect microcirculatory perfusion. Unfortunately, no technology is available nowadays for the measurement of tissue PCO 2 .
The clinical use of P cv-a CO 2 /C a-cv O 2 as a substitute for RQ is conflictive. First, the increase in RQ secondary to critical reductions in DO 2 is a life-threatening and striking condition. It is an easily noticeable event, which does not probably require further monitoring. Given that the start of anaerobic metabolism is indicated by the sharp rise in the RQ, and the normal range of RQ is wide, the use of a defined cutoff of 1.4 for P cv-a CO 2 /C a-cv O 2 is irrelevant. Moreover, P cv-a CO 2 /C a-cv O 2 is more dependent on factors that modify the CO 2 Hb dissociation curve than on the actual RQ. Experimental studies showed that RQ and P cv-a CO 2 /C a-cv O 2 might exhibit distinct behaviors in different models. The ability of P cv-a CO 2 /C a-cv O 2 to predict the mortality of critically ill patients is not superior, but probably lower than that of lactate. In addition, the association with mortality could be related to the impact of acidosis and anemia on the ratio. Regardless of its meaning, the relationship of P cv-a CO 2 /C a-cv O 2 to oxygen supply dependency is controversial. A randomized controlled trial also showed that P cv-a CO 2 /C a-cv O 2 is useless as a goal of resuscitation in sepsis. The use of P cv-a CO 2 /C a-cv O 2 as an index of tissue oxygenation lacks a physiological basis and solid evidence.
In brief, P cv-a CO 2 and P cv-a CO 2 /C a-cv O 2 are complex variables with multiple determinants. Accordingly, their interpretation requires careful analysis. The direct assumption that high values of P cv-a CO 2 and P cv-a CO 2 /C a-cv O 2 are signs of microcirculatory hypoperfusion and anaerobic metabolism should be avoided. P cv-a CO 2 is a marker of cardiac output. In states of low cardiac output, increased P cv-a CO 2 reflects global tissue hypoperfusion. In conditions of normal or high cardiac output, high values should be explained by changes in the two other determinants, the CO 2 Hb dissociation curve and the VCO 2 , and not by an altered microcirculation. Since the calculation of P cv-a CO 2 /C a-cv O 2 is derived from the determinants of the RQ, it has been considered a surrogate for RQ and tissue oxygenation. Nevertheless, it is more dependent on factors that modify the dissociation of CO 2 from Hb than on the actual RQ measured by analysis of expired gases. Therefore, high values should be interpreted with extreme caution.