ASSOCIATION BETWEEN THE OXYGEN CONSUMPTION: LACTATE RATIO AND SURVIVAL IN CRITICALLY ILL PATIENTS WITH SEPSIS

Abstract

Introduction:

Mitochondrial dysfunction leading to impairment of oxygen extraction, referred to as cytopathic hypoxia, contributes to morbidity in sepsis. Oxygen consumption (VO₂) may be a useful measure of the severity of cytopathic hypoxia. We monitored VO₂ and carbon dioxide production (VCO₂) in septic patients and investigated the association with hospital survival.

Methods:

We retrospectively identified adult (≥18 years) septic patients from a larger prospective observational cohort of critically ill patients on mechanical ventilation. A gas-exchange monitor recorded continuous VO₂ and VCO₂ for up to 48 h. We then tested the association of median VO₂, VCO₂, respiratory quotient (RQ), and the VO₂:lactate ratio with survival.

Results:

A total of 46 septic patients were included in the analysis, of whom 28 (61%) survived. Overall median VO₂ was not associated with survival (3.72 mL/kg/min [IQR: 3.39, 4.92] in survivors and 3.42 mL/kg/min [IQR: 2.97, 5.26] in non-survivors, P = 0.12). The overall median VCO₂ and RQ were also not associated with survival. Adjusting for age and the presence of shock did not change these results. The VO₂:lactate ratio was associated with survival (adjusted OR 2.17 [95% CI 1.12, 4.22] per unit increase in ratio, P = 0.03). The percent change in median VCO₂ was 11.6% [IQR: −8.2, 28.7] in survivors compared with −8.3% [IQR: −18.0, 4.7] in non-survivors (P = 0.03). The percent changes in median VO₂ and RQ were not different between groups.

Conclusion:

The VO₂:lactate ratio was significantly higher in survivors, while there was no association between median VO₂ alone and survival. There was a significant difference in change in VCO₂ over time between survivors and non-survivors.

INTRODUCTION

Sepsis is a common cause of hospital admission in the United States (1) and is the leading cause of death in non-cardiac intensive care units (2, 3). Septic shock is characterized by evidence of end-organ damage and elevation in lactate. Historically, the main explanation given for this derangement was inadequate oxygen delivery due to tissue hypoperfusion. In recent years, however, there is growing evidence that cytopathic hypoxia (impaired oxygen extraction from mitochondrial injury) contributes to hyperlactatemia and organ dysfunction in sepsis (4–7). Oxygen consumption (VO₂), which is determined by both oxygen delivery and oxygen extraction, may be a useful measure of the degree of impairment in oxygen extraction, and low VO₂ has been associated with mortality in septic shock and after cardiac arrest (8–10).

This area of research has been limited by the difficulty of measuring VO₂ reliably and accurately in the critically ill (11, 12). In earlier studies, VO₂ was either calculated, which required placement of a pulmonary artery catheter, or measured by indirect calorimetry, using large and somewhat cumbersome equipment that is not available in many intensive care units (13–15). Newer technology allows for direct, non-invasive, and continuous measurement of VO₂ and carbon dioxide production (VCO₂) in patients who are mechanically ventilated (16, 17). Our group has used this technology previously to evaluate whether VO₂ is associated with survival after cardiac arrest (10). The same study also explored the ratio between VO₂ and lactate as a possible more sensitive marker of the adequacy of VO₂ for a specific patient at a specific moment and found a stronger association with mortality using this ratio than using VO₂ alone. Oxygen demand can vary by patient due to temperature, levels of sedation, disease state, and other factors (17, 18). A lower VO₂ may not therefore always be pathologic or indicative of harm. The VO₂:lactate ratio accounts for this, but has previously only been tested in a small postarrest cohort (10).

In the present study, we monitored VO₂ and VCO₂ continuously for up to 48 h in a cohort of septic patients to investigate whether VO₂ and the ratio of VO₂ to lactate were associated with hospital survival in patients with sepsis. We hypothesized that a higher median VO₂ and VO₂:lactate ratio and an increase in VO₂ over time would be associated with hospital survival. We secondarily assessed the association between VCO₂ and the respiratory quotient (RQ) with hospital survival. Some of the results of this study have been previously reported in the form of an abstract (19).

METHODS

Design and setting

This was an observational study conducted at Beth Israel Deaconess Medical Center, an urban tertiary care center in Boston, Mass, from January 2016 through March 2019. In this study, we retrospectively identified septic patients who were part of an ongoing prospective observational cohort of oxygen metabolism in critically ill adult patients receiving mechanical ventilation. This study was approved by the local Institutional Review Board. All legally authorized surrogates provided verbal consent prior to enrollment in the study.

Study population and outcomes

We included adult patients (age ≥ 18 years) admitted to the medical or surgical intensive care unit with a diagnosis of sepsis who required mechanical ventilation and had stable ventilator settings for at least 3 h prior to enrollment. Patients were excluded if they had: factors known to alter VO₂ such as air-leak (e.g., chest tube) or agitation, positive end-expiratory pressure >12 cm H₂O, due to the potential risk of a brief disconnection from the ventilator, fraction of inspired oxygen (FiO₂) >60% due to the monitor being validated for FiO₂ of 60% or less, and/or anticipated extubation within 24 h of the enrollment. Patients with less than 12 h of data were excluded prior to analysis.

The primary outcome was hospital survival. Sepsis was defined as a suspected or confirmed bacterial infection with evidence of a systemic inflammatory response for which the patient was receiving antibiotics at the time of enrollment. Septic shock was defined as meeting these same criteria with the addition of shock, defined as requiring vasopressor support (norepinephrine, epinephrine, vasopressin, and/or phenylephrine) for at least 30 min. The diagnosis of sepsis was based on medical chart review.

VO₂ and VCO₂ measurement

Continuous VO₂ and VCO₂ measurements were obtained with the CARESCAPE B650 Monitor with Respiratory Module E-sCOVX (GE Healthcare, Helsinki, Finland). This monitor has been approved for the measurement of VO₂ and VCO₂ in critically ill mechanically ventilated patients and has been validated against indirect calorimetry using the metabolic cart (15). The CARESCAPE monitor connects in-line with the patient’s ventilator tubing and has a built-in module for measuring spirometry and gas exchange (Fig. 1). Via gas sampling ports and a flow sensor connected in-line with the ventilator tubing, the gas exchange module measures the flow and volume of exhaled gas and the difference in O₂ and CO₂ content between inhalation and exhalation using a pneumotachograph and a rapid paramagnetic analyzer. The monitor provides continuous readings with each patient breath. The GE S/5 Collect software records and saves all values averaged over a chosen interval that, for this study, was every minute.

FIG. 1.

Representation of the d-lite (translucent yellow tubing) with gas sampling line, which attaches in-line to the ventilator tubing between the endotracheal tube and the Y-connector, for gas sampling for metabolic measurements.

Data collection

The CARESCAPE monitor was connected to the patient’s ventilator tubing and set to record data for up to 48 h. In addition to respiratory parameters, we abstracted clinical data on each patient from the electronic medical record, which was saved in an internal database. These data included demographics, past medical history, ventilator settings, lactates, usage of neuromuscular blocking agents, vasopressors, and/or sedatives, and hospital survival.

Statistical analysis

We used descriptive statistics to present baseline characteristics. Continuous variables were summarized using mean ± standard deviation or median and interquartile range (IQR), with differences between groups tested using Student t test or Wilcoxon rank-sum test, depending on the normality of the data. Categorical variables were summarized by frequencies and percentages with differences between groups tested using Chi-square or Fisher exact test, as applicable.

Prior to analysis, metabolic data was cleaned using an algorithm designed by our research team in R version 3.51 (R Foundation for Statistical Computing, Vienna, Austria). This automated algorithm excluded VO₂ and VCO₂ data if one or more of the following criteria was met: all values recorded in the 10 min following a change in FiO₂ of ≥10%, all values recorded while FiO₂ was >60%, all values deviating ≥20% from the preceding value and more than two standard deviations away from the mean of the next five values, all values more than two standard deviations away from the mean of the corresponding 60-min interval, and all values out of physiologic range unless these were persistent for more than 30 min, as these measurements were considered artifacts. Values considered out of physiologic range, based on the existing literature (8, 20), were VO₂ <100 mL/min or >1,000 mL/min and VCO₂ <70 mL/min or >800 mL/min. We allowed these values if they persisted for at least 30 min and if they were not excluded per the algorithm for other reasons, as critically ill patients can sometimes have values well outside of standard normal range, due to medication use such as sedatives and neuromuscular blockade and/or alterations in metabolic function. For one patient, we manually excluded values of VCO₂ that were severely out of physiologic range (>2,000 mL/kg/min) as these were attributed to monitor error. Concurrent values of VO₂ and RQ for that patient were also dropped as only time points with valid measures of all three parameters were included. See Figure 2 for examples of algorithm performance. RQ was defined as VCO₂ divided by VO₂. RQ values were calculated where both VCO₂ and VO₂ measurements were available within the same minute. After data cleaning, all VO₂ and VCO₂ data were adjusted for patient weight in kilograms (mL/kg/min).

FIG. 2. Figures (A) to (C) illustrate how our automated R algorithm performed and cleaned data from oxygen consumption (VO₂ depicted in mL/min) based on 48 h of recordings from three different individual subjects.

The black circles represent fraction of inspired oxygen (FiO₂) in percentage. The gray circles represent the excluded VO₂ data. The red circles represent the VO₂ data used for the analysis. Gray columns represent 60-minute intervals.

We calculated the overall median VO₂ and VCO₂ for the study population based on the individual median of all time points. We also computed patient medians during the first and last 5 h of data collection. We planned to categorize the relative change in median VO₂, VCO₂, and RQ from the first 5 h to the last 5 h of data collection into three trend groups (increasing, decreasing, or constant), with a change threshold of 20%. Calculation of the VO₂:lactate ratio was based on the first lactate drawn after enrollment and the VO₂ within the same minute. Two logistic regression models were used to assess the relationships between overall median VO₂ and VO₂:lactate ratio to survival, adjusting for age (continuous covariate) and septic shock status (binary covariate). Due to the pilot nature of this study, no formal power analysis was performed.

All tests were two-sided and the nominal level of statistical significance (α) was 5%. Analyses were performed using R version 3.51 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Patient characteristics

A total of 48 septic patients were enrolled. Two were excluded, one due to withdrawal from the study and one due to monitor dysfunction resulting in no usable metabolic data (Fig. 3). We included 46 patients in the analysis, of whom 28 (61%) survived and 33 (72%) were in septic shock at the time of enrollment. Baseline characteristics in survivors and non-survivors are presented in Table 1. Time from intubation to enrollment (start of metabolic data collection) ranged from 3 to 47 h, with a median of 9 h [IQR: 8, 12].

FIG. 3. A flow diagram to illustrate the selection process of septic patients.

We consented 168 patients for a larger prospective observational study from which we retrospectively identified 48 septic patients. Two of these were excluded and 46 were included in the analysis. *Consented not enrolled includes patients of which their legally authorized representative consented, but the patient was not enrolled due to either alternations of ventilator settings or change in goals of care prior to enrollment, ⁺Sepsis was defined as a suspected or confirmed bacterial infection with evidence of a systemic inflammatory response, for which a patient was receiving antibiotics during time period of enrollment, ^‡ No measurements were saved.

Table 1.

Baseline characteristics stratified by survival status

	All (n = 46)	Survivors (n = 28)	Non-survivors (n = 18)
Age (years), mean ± SD	64 ± 14	62 ± 14	69 ± 13
Female, n (%)	24 (52)	16 (57)	8 (44)
Race/ethnicity , n (%)
Black	4 (9)	3 (11)	1 (6)
White	37 (80)	22 (79)	15 (83)
Other	5 (11)	3 (11)	2 (11)
Past medical history, n (%)
CAD	9 (20)	6 (22)	3 (17)
Cancer	10 (22)	3 (11)	7 (39)
CHF	4 (9)	3 (11)	1 (6)
COPD	12 (27)	7 (26)	5 (28)
Diabetes	18 (40)	11 (41)	7 (39)
Hypertension	24 (53)	12 (44)	12 (67)
Liver disease	6 (13)	2 (7)	4 (22)
Renal disease	7 (16)	2 (7)	5 (28)
Stroke	7 (16)	3 (11)	4 (22)
ICU settings
Ventilator settings, median (IQR)	5 (5, 8)	8 (5, 9)	5 (5, 6)
PEEP, cm H2O	40 (40, 50)	40 (40, 50)	40 (40, 50)
FiO2, %
Medication at enrollment, n (%)	33 (72)	18 (64)	15 (83)
Vasopressor*	45 (98)	28 (100)	17 (94)
Sedation†	2 (4)	2 (7)	0 (0)
Paralytics‡
Medication during enrollment, n (%)	39 (85)	23 (82)	16 (89)
Vasopressor*	46 (100)	28 (100)	18 (100)
Sedation†	6 (13)	5 (18)	1 (6)
Continuous paralytics‡
Lactate
Initial lactate, mmol/L, median (IQR)	1.9 (1.4, 3.0)	1.8 (1.2, 2.7)	2.1 (1.6, 4.8)
Initial lactate ≥ 2 mmol/L, n (%)	19 (48)	10 (42)	9 (56)
Location of infection, n (%)
Pulmonary	29 (63)	20 (71)	9 (50)
Urinary tract	11 (24)	6 (21)	5 (28)
Brain/meninges	3 (7)	1 (4)	2 (11)
Bloodstream	10 (22)	8 (29)	2 (11)
Soft tissue	3 (7)	1 (4)	2 (11)
Intraabdominal	8 (17)	6 (21)	2 (11)
Other§	4 (9)	3 (11)	1 (6)

^*Vasopressors included norepinephrine, vasopressin, phenylephrine, and epinephrine.

^†Sedation included fentanyl, midazolam, propofol, and dexmedetomidine.

^‡Paralytics included vecuronium, rocuronium, cisatracurium.

^§Other locations included one in the endometrium and three were unknown at the time of study period.

CAD indicates coronary artery disease; CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; FiO₂, fraction of inspired oxygen; IQR, interquartile range; PEEP, positive end-expiratory pressure; SD, standard deviation.

Oxygen consumption (VO₂)

In unadjusted analysis, there was no association between overall median VO₂ and survival (3.72 mL/kg/min [IQR: 3.39, 4.92] in survivors and 3.42 mL/kg/min [IQR: 2.97, 5.26] in non-survivors, P = 0.12). This remained unchanged when adjusted for age and presence of shock (adjusted odds ratio [aOR] for survival 1.46 per 1 mL/kg/min increase in median VO₂ [95% CI: 0.88, 2.42], P = 0.14).

VO₂:lactate ratio

At least one lactate measurement was available for 38 of 46 (83%) patients, of which 33 (87%) were drawn within 24 h of study start. Using a Wilcoxon ranked sum test, there was no significant difference in lactate between survivors and non-survivors (P = 0.26). A higher initial VO₂:lactate ratio was associated with survival (median 3.28 [IQR: 2.05, 3.62] in survivors and 1.41 [IQR: 0.85, 2.39] in non-survivors, P = 0.03) (Fig. 4). The relationship remained statistically significant when adjusted for age and presence of shock (aOR 2.17 [95% CI: 1.12, 4.22] per unit increase in ratio, P = 0.03). The area under the receiver operating curve (AUROC) was 0.76 [95% CI: 0.59, 0.93] for the VO₂:lactate ratio, compared with 0.62 [95% CI: 0.43, 0.82] for VO₂ alone and 0.65 [95% CI: 0.46, 0.85] for lactate alone. The difference between the AUROC for VO₂:lactate and lactate alone (P = 0.17) or VO₂ alone (P = 0.19) was not statistically significant.

FIG. 4. A box plot to illustrate the median initial VO₂:lactate ratio in survivors versus non-survivors.

The middle bar in the box represents the median. The lower and upper bars represent the 25th percentile and the 75th percentile, respectively. VO₂ indicates oxygen consumption.

RQ and VCO₂

There was no relationship between RQ or VCO₂ and hospital survival. The overall median VCO₂ in survivors was 2.77 mL/kg/min [IQR: 2.38, 3.38] compared with 2.43 mL/kg/min [IQR: 2.14, 3.38] in non-survivors (P = 0.19). The overall median RQ in survivors was 0.69 [IQR: 0.68, 0.72] compared with an RQ of 0.69 [IQR: 0.66, 0.73] in non-survivors, P = 0.78.

Trends over time

We were unable to run the planned analysis comparing directions of change within 48 h due to the small sample size in some of the trend groups. We therefore performed a post hoc analysis comparing the percent change in VO₂, VCO₂, and RQ from the first 5 h to the last 5 h in survivors and non-survivors. There was a significant difference in percent change in median VCO₂ (11.6% [IQR: −8.2, 28.7] in survivors and −8.3% [IQR: −18.0, 4.7] in non-survivors, P = 0.03). Median VCO₂ over time in survivors and non-survivors is presented in Figure 5. The percent change in median VO₂ (−3.0% [IQR: −11.0, 17.7] in survivors and −3.1% [IQR: −15.7, 4.4] in non-survivors, P = 0.27) and RQ (2.3% [IQR: −2.4, 8.0] in survivors vs. 1.58% [IQR: −5.3, 6.6] in non-survivors, P = 0.46) was not significantly different between groups.

FIG. 5. Median of all individual median VCO₂ values by hour in survivors (beige triangles) and non-survivors (blue dots).

Lines and bars indicate interquartile range. VCO₂ indicates carbon dioxide production.

DISCUSSION

In this study, we investigated the relationship between VO₂, the VO₂:lactate ratio and survival in septic adults, secondarily assessing whether VCO₂ and RQ were associated with survival. We found that the VO₂:lactate ratio was significantly higher in survivors, while there was no association between median VO₂ alone and survival. We also found a significant difference in change in VCO₂ over time between survivors and non-survivors but did not find any difference in change over time in VO₂ or RQ.

The understanding of the role of oxygen delivery and metabolism in critical illness continues to evolve. Early studies found that increasing oxygen delivery by augmenting cardiac output and the oxygen carrying capacity of the blood increased VO₂ and was beneficial in terms of shorter length of stay in the intensive care unit and better survival (9, 21, 22). Subsequent studies, however, found that only some patients were able to increase VO₂ in response to increased delivery, and those who were unable to increase VO₂ had higher hospital mortality (7, 20). A study by Hayes et al. compared the oxygen delivery and consumption pattern of 78 sepsis survivors and non-survivors receiving dobutamine to target supratherapeutic levels of oxygen delivery (DO₂). They demonstrated that non-survivors, despite increases in DO₂, were unable to increase VO₂, and had higher lactates and higher mortality (7). This was one of several studies (4, 6, 23–25) that led to the current prevailing hypothesis that oxygen extraction is of critical importance in the prognosis of septic shock. The importance of oxygen extraction was also demonstrated in a recent study by Gattinoni et al. (5). Those investigators plotted the relationship between central venous saturation (ScvO₂) and lactate, finding that lactate and mortality are high in patients with both very low and very high ScvO₂. They concluded that the high-lactate, high ScvO₂ pattern in sepsis, caused by impaired oxygen extraction, is a major driver of mortality. Findings like these support the importance of cytopathic hypoxia and highlight the potential harms of focusing solely on hypoperfusion when designing interventions for septic shock (20). There is currently no proven intervention to improve oxygen extraction, but a reliable method to monitor this parameter could be prognostically useful and provide a potential target for intervention.

VO₂ is determined by delivery and extraction, both of which can be significantly altered in sepsis: The oxygen “supply” may vary somewhat due to alterations in cardiac output or arterial oxygen content (delivery) and/or decreases in extraction ability due to mitochondrial dysfunction. The oxygen “demand” may also vary depending on a patient’s native metabolic rate as well as external factors such as sedation, neuromuscular blockade, temperature, or disease state. Thus, a low VO₂ may not always be indicative of harm, as long as the level is adequate to meet a patient’s needs. The adequacy of a given number for VO₂ may be best interpreted in conjunction with lactate or some other marker of tissue oxygenation. We propose using the VO₂:lactate ratio to provide a more precise metabolic picture and to differentiate between a patient whose VO₂ is appropriately low (low VO₂ and low lactate) and one whose VO₂ is inappropriately low and likely to be harmful (low VO₂ and high lactate). Our group recently published a small study in post-cardiac arrest patients demonstrating that higher VO₂:lactate ratio was associated with survival, similar to our findings in the present study (10). The lack of a significant relationship between VO₂ and survival in the current cohort, combined with the positive association of the VO₂:lactate ratio with survival, suggests that perhaps VO₂ is not a sufficient predictor of survival in sepsis when used as an individual variable but is useful when combined with a measure of its adequacy for that specific patient. The VO₂:lactate ratio also appeared to have a stronger association with survival than lactate alone, although this difference did not achieve statistical significance.

The significance of the difference in change in VCO₂ over time between survivors and non-survivors is unclear. One hypothesis, which has been suggested in post-cardiac arrest patients and may be similar in sepsis, is that non-survivors may have more severe mitochondrial injury and thus be more dependent on extramitochondrial respiration, which does not lead to the same CO₂ production per unit of oxygen consumed (10, 26). Especially in light of the lack of a concurrent difference over time in VO₂ and RQ however, the isolated finding in VCO₂ should be validated in a larger cohort before conclusions can be drawn.

Our study has several limitations. First, the diagnosis of sepsis was determined retrospectively based on review of medical charts. The timeline of sepsis is very important in that there may be different stages of metabolic dysfunction at different time points (27), which we were not able to account for. This was an observational study, and therefore we did not have control of when or if the clinical team drew a lactate. This may affect our VO₂:lactate ratio, since the association of this variable with outcome may also depend on at what time point it is calculated. In addition, 18 of the 46 patients (39%) received at least one dose of intravenous thiamine during the study period, and limited data suggests thiamine may have an effect on VO₂ (28, 29). Some recent papers have questioned the use of the respiratory module E-sCOVX, stating that the measurements are overestimated (16, 17). However, the over-estimation was found to be approximately 10%. As we were more interested in relative differences in VO₂ than absolute values, this small overestimation, if present, likely would not change our conclusions. VO₂ measurements in critically ill patients are also frequently affected by artifact from things such as changes in FiO₂ or medications. We have tested our algorithm to reduce inclusion of artifactual values in multiple patients and it appears to perform well (Fig. 2), but this risk cannot yet be eliminated completely. Finally, our sample size was small, and our study was therefore not adequately powered to detect small differences in metabolic parameters between groups. Finally, due to the small sample size these findings should be considered preliminary and validated in future studies in larger cohorts and at other institutions. Despite these limitations, this study contributes to a better understanding of the potential of VO₂ as a prognostic value in critically ill patients with sepsis measured by an easy-to-use bedside monitor.

CONCLUSIONS

In conclusion, we found that the ratio of VO₂:lactate was associated with survival in a small cohort of septic, mechanically ventilated patients, while median VO₂ alone was not associated with survival. Whether the VO₂:lactate ratio could be a useful treatment target or prognostic tool warrants further investigation.