The main findings of our study were that in fluid-responder septic shock patients (1) the ability of ∆ContCO2/∆ContO2 and ∆PCO2/∆ContO2 ratios to predict the presence of global anaerobic metabolism was excellent and higher than lactate; (2) ∆ContCO2/∆ContO2 ratio was not a better marker of global anaerobic metabolism than ∆PCO2/∆ContO2 ratio; (3) ScvO2 failed to detect the presence of global tissue hypoxia, except for values ≥80 %.
When DO2 is acutely reduced, VO2 remains stable (oxygen supply independency) because the tissues adapt their OE proportionally. When DO2 falls below a critical low value, a proportionate increase in OE cannot be maintained and the VO2 starts to fall (oxygen supply dependency) and tissue hypoxia occurs as reflected by an abrupt increase in blood lactate concentration [13–15]. Thus, VO2/DO2 dependence has been considered to be a hallmark of tissue hypoxia and the activation of anaerobic metabolism [14–16], although it has been challenged because of the methodological limitations (mathematical coupling) in the VO2/DO2 relationship assessment [17, 18].
We defined an increase in VO2 ≥ 15 % as a clinically significant augmentation by similarity to the definition of the increase in cardiac index, since VO2 is proportional to this variable. This cutoff value was chosen by the fact that the least significant change in CI measured by transpulmonary thermodilution is 11.0 % when 20 mL is used to perform iced saline injections in triplicate (unpublished data). On the other hand, this definition of “VO2 response” allows us comparing our findings with those of a previous study [8]. We confirm the results of Monnet et al. [8] that calculating VO2 from the central instead of the mixed venous blood also allows to detect the presence of anaerobic metabolism through VO2/DO2 dependence. Indeed, in our study, the baseline lactate concentration was elevated in patients with VO2/DO2 dependency phenomenon and higher than in patients with VO2/DO2 independency (Table 2). Moreover, it is hard to believe that mathematical coupling of measurement errors was responsible for the VO2/DO2 dependency in our study. Indeed, we observed that VO2/DO2 dependency occurred in one subgroup of patients but not in others, despite similar changes in DO2 (Table 2). Such a methodological problem can hardly account for the existence of VO2/DO2 dependency only in one subgroup. If that were an issue, one would expect it to influence results uniformly. Finally, the increase in VO2 could have resulted from an additional non-mitochondrial non-oxidative oxygen uptake when dysoxia has resolved [19]. However, this mechanism is less likely to have occurred in our study because the observed mean slope of the VO2/DO2 relationship in the subgroup of VO2-responders was 47.8 ± 10 %, suggesting VO2/DO2 dependency and activation of anaerobic metabolism (Table 2) [20].
In experimental conditions of tissue hypoxia, a smaller reduction in VCO2 than VO2 has been observed, suggesting a non-aerobic production of CO2 [21–23]. Therefore, the occurrence of a high RQ may be considered as a sign of anaerobic metabolism. Recently, Monnet et al. [8] have used the ∆PCO2/∆ContO2 ratio as a surrogate of RQ and found that a ∆PCO2/∆ContO2 ratio at baseline ≥ 1.8 mmHg/mL predicted accurately VO2/DO2 dependence among patients whose DO2 increased after fluid administration. Our results agree with those findings, and interestingly, the observed cutoff value of this ratio, in our septic shock patients, was almost similar to what was found in the Monnet et al. report [8].
The use of ∆PCO2/∆ContO2 ratio as a surrogate of VCO2/VO2 assumes that the PCO2/CO2 content relationship is quasi-linear, which may be true over the physiologic range of PCO2 [24]. However, this relationship can be affected by the degree of metabolic acidosis, hematocrit, and oxygen saturation (Haldane effect), and it becomes nonlinear if these factors change [25]. In this regard, it has been shown that venous-to-arterial PCO2 differences and venous-to-arterial CO2 content differences might change in opposite direction in splanchnic region under conditions of very low venous oxygen saturation [26]. Recently, Ospina-Tascon et al. found a significant association with mortality, in septic shock patients, for the mixed ∆ContCO2/mixed ∆ContO2 ratio but not for the mixed ∆PCO2/mixed ∆ContO2 ratio [27]. Thus, ∆ContCO2/∆ContO2 ratio could be a more reliable marker of tissue hypoxia than ∆PCO2/∆ContO2. However, we found that ∆ContCO2/∆ContO2 was not better predictor of tissue hypoxia than the ∆PCO2/∆ContO2 in septic shock patients (Fig. 2). It does not seem that Haldane effect has played an important role in our study. Furthermore, the degree of metabolic acidosis, as reflected by base excess, was not severe enough to significantly affect the PCO2/CO2 content relationship in our septic shock patients. Even though the ∆ContCO2/∆ContO2 ratio more physiologically mirrors RQ compared with ∆PCO2/∆ContO2, we found that both ratios can be used accurately to predict fluid responsiveness at tissue level. However, the computation of ∆PCO2/∆ContO2 ratio is less cumbersome and less subject to the risk of errors, and therefore, it is much easier to be used at the bedside.
Lactate value was not good to detect the presence of anaerobic metabolism in our septic shock patients. Our results are in discrepancy with previous findings [7, 8, 28]. However, hyperlactatemia does not necessarily reflect anaerobic metabolism secondary to tissue hypoxia, especially in septic states [3, 4, 29]. Other non-hypoxic mechanisms such as accelerated aerobic glycolysis induced by sepsis-associated inflammation [30], inhibition of pyruvate dehydrogenase [31], and impaired lactate clearance [32] may contribute to hyperlactatemia found in septic patients. In endotoxic states, lactate levels failed to discriminate between hypoxia and aerobiosis [33]. Furthermore, Rimachi et al. [34] found the presence of hyperlactatemia in 65 % of septic shock patients, but only 76 % of these patients also had a high lactate/pyruvate ratio confirming the non-hypoxic cause of hyperlactatemia in septic states. Moreover, in fluid-responder patients, we found no significant relationship between changes in VO2 induced by VE and changes in lactate levels, whereas changes in ∆PCO2/∆ContO2 and ∆ContCO2/∆ContO2 ratios were correlated well with changes in VO2. This finding suggests that these ratios respond to changes in global tissue oxygenation faster than blood lactate concentration likely due to the alteration of lactate clearance.
The majority (71 %) of our septic shock patients had a ScvO2 value ≤70 % at their inclusion in the study. This finding is due to the fact that patients, in our study, were recruited in the very early period of acute circulatory failure; the time between the start of care and enrollment was only 102 min. Within this period, septic shock patients are not fully resuscitated yet, and as a consequence, low values of ScvO2 are observed more frequently [35–37]. Even though our population seems to be different from that in the study of Monnet et al. [8], we confirm that ScvO2 is a poor predictor of the presence of anaerobic metabolism. This can be explained by the fact that ScvO2 is not a regulated parameter but an adaptive one that depends on four regulated constituents: oxygen consumption, hemoglobin, SaO2, and cardiac output. Therefore, ScvO2 is widely fluctuating. However, these results should not dissuade us from monitoring ScvO2 but encourage us to include it in a multimodal approach. Indeed, a low ScvO2 value reflects the inadequacy of oxygen supply, and fluid administration can be helpful in order to correct oxygen supply/demand imbalance, even in situations of VO2/DO2 independency, to avoid further decreases in DO2 below a critical value leading to tissue hypoxia. On the other hand, only ScvO2 values ≥80 % were able to predict the presence of global tissue hypoxia with a high specificity. All these patients also had higher lactate levels and higher ∆PCO2/∆ContO2 and ∆ContCO2/∆ContO2 ratios. This suggests that these patients had a greater alteration of their microcirculation due to sepsis than the other fluid-responder patients. However, this finding should be interpreted with caution, since only six patients had a baseline ScvO2 value ≥80 % in fluid-responders’ group, and our study was not designed for testing this hypothesis.
Contrary to what was found previously [8], DO2 did not decrease during VE, in fluid-non-responder patients, even though arterial hemoglobin significantly decreased by 5.6 ± 4.6 % (Additional file 2: Table S1), which was lower than that in the Monnet et al. study (8 ± 4 %) [8]. The discrepancy between the two studies may be due to dissimilar populations of patients and to the differences in the time to inclusion, which was longer in the study by Monnet et al. [8] than that in our study (6.1 vs. 1.7 h) explaining the more pronounced hemodilution effect in their study.
We believe our findings add significant values to the Monnet et al. study [8]. Indeed, we have demonstrated that the ∆PCO2/∆ContO2 ratio is a reliable marker of global anaerobic metabolism in the very early period of septic shock where patients are still not fully resuscitated and that ∆ContCO2/∆ContO2 is not superior to ∆PCO2/∆ContO2 for predicting the presence of global tissue hypoxia in these patients. Furthermore, our study shed the light on the fact that hyperlactatemia should not always be regarded as reflecting the presence of global tissue hypoxia, especially in septic shock patients. This finding is of clinical importance since elevated lactate values could incite the clinician to undertake unnecessary interventions, with their potentially harmful effects, such as tissue edema and positive fluid balance, which have constantly been associated with poorer outcome [38].
Our study presents several limitations. First, we used central venous blood instead of mixed venous to assess VO2- and CO2-derived variables, and thus, we might have missed the evaluation of the splanchnic oxygenation. However, our study is the second one that confirms that calculating oxygen-derived variables from the central venous blood is able to detect the presence of tissue hypoxia through VO2/DO2 dependence. Second, regional or local tissue hypoxia might not be detected by the assessment of the changes in global oxygen consumption. Finally, it is a single-center study, which may limit its external validity.