Veno‐arterial CO2 difference and lactate for prediction of early mortality after cardiac arrest

Patients admitted to intensive care after cardiac arrest are at risk of circulatory shock and early mortality due to cardiovascular failure. The aim of this study was to evaluate the ability of the veno‐arterial pCO2 difference (∆pCO2; central venous CO2 – arterial CO2) and lactate to predict early mortality in postcardiac arrest patients. This was a pre‐planned prospective observational sub‐study of the target temperature management 2 trial. The sub‐study patients were included at five Swedish sites. Repeated measurements of ∆pCO2 and lactate were conducted at 4, 8, 12, 16, 24, 48, and 72 h after randomization. We assessed the association between each marker and 96‐h mortality and their prognostic value for 96‐h mortality. One hundred sixty‐three patients were included in the analysis. Mortality at 96 h was 17%. During the initial 24 h, there was no difference in ∆pCO2 levels between 96‐h survivors and non‐survivors. ∆pCO2 measured at 4 h was associated with an increased risk of death within 96 h (adjusted odds ratio: 1.15; 95% confidence interval [CI]: 1.02–1.29; p = .018). Lactate levels were associated with poor outcome over multiple measurements. The area under the receiving operating curve to predict death within 96 h was 0.59 (95% CI: 0.48–0.74) and 0.82 (95% CI: 0.72–0.92) for ∆pCO2 and lactate, respectively. Our results do not support the use of ∆pCO2 to identify patients with early mortality in the postresuscitation phase. In contrast, non‐survivors demonstrated higher lactate levels in the initial phase and lactate identified patients with early mortality with moderate accuracy.


| INTRODUCTION
Hospital mortality for patients admitted after an out-of-hospital cardiac arrest (OHCA) is approximately 50%. 1 Although most deaths are secondary to hypoxic-ischemic brain injury and subsequent withdrawal of life-sustaining therapy (WLST), a significant number of patients die due to cardiovascular and/or multiorgan failure. 2,3 After return of spontaneous circulation (ROSC), patients are at risk of myocardial depression, 4 systemic inflammatory response syndrome with endothelial dysfunction and vasodilatation 5 and microcirculatory dysfunction 6,7 that may result in occult or overt circulatory shock with organ failure and an increased risk of in-hospital death. 5,[8][9][10] Early, non-neurologic death typically occurs earlier than death secondary to hypoxicischemic brain injury, often within the first 48 h after ROSC. 3 Identifying patients at risk may be helpful to guide resuscitation efforts.
Lactate is used to assess circulatory failure in critically ill patients.
In postcardiac arrest patients, a higher lactate at admission and a low lactate clearance during the initial 24 h have been associated with lower survival. [11][12][13][14][15] Lactate elevation in the initial postresuscitation phase correlates with the duration of resuscitation and lactate levels are higher in patients treated with mild therapeutic hypothermia. 16,17 Consequently, the predicative value of lactate to identify ongoing hypoperfusion in postcardiac arrest patients may be limited and other markers of insufficient oxygen delivery could be of value.
Veno-arterial carbon-dioxide difference (ΔpCO 2 ) is considered a marker of tissue perfusion. 18,19 Elevated ΔpCO 2 is associated with low cardiac output, 20 impaired microcirculation 21,22 and has been linked to a poor outcome in patients with septic shock 23,24 and in postoperative patients. 25 ΔpCO 2 may be a more appropriate marker than lactate to identify patients at risk of mortality in the early postresuscitation phase as it is rapidly normalized after ROSC and therefore may reflect real-time tissue perfusion in postcardiac arrest patients with greater precision. 26 Data on ΔpCO 2 in postcardiac arrest patients is limited. 27,28 In this target temperature management 2 (TTM2) sub-study, we hypothesized that patients at risk of early in-hospital mortality would have hypoperfusion that could be identified by an elevated ΔpCO 2 and that the prognostic characteristics of ΔpCO 2 would potentially be superior to lactate.

| METHODS
This was a pre-planned sub-study of the Targeted Hypothermia versus Targeted Normothermia after OHCA (TTM2) trial and involved five Swedish intensive care units that were recruiting patients to the TTM2 trial. The TTM2 trial compared two different temperature regimens in comatose patients after OHCA. Patients were randomized to hypothermia at 33 C or normothermia with treatment of fever (≥37.8 C). The primary outcome was all-cause mortality at 6 months.
The results of the TTM2 trial have been published. The current substudy was approved by the Regional Ethical Review Board in Lund (Dnr: 2018/832).

| Patients
Patients included in the main trial and admitted to any of the intensive care units involved in this sub-study were eligible for inclusion. The eligibility criteria in the TTM2 trial were unconscious patients ≥18 years of age admitted for intensive care after OHCA with a presumed cardiac etiology. To be included in the sub-study, patients had to have a central venous catheter (CVC) inserted within 4 h after randomization, recorded data of ΔpCO 2 , and known survival status at 96 h after intensive care unit (ICU) admission. Exclusion criteria were consistent with the TTM2 protocol. 29

| Patient management
Patients were managed according to the TTM2 study protocol. 29 In brief, patients were randomized either to treatment with hypothermia at 33 C or to active temperature control (if the temperature increased above 37.7 C). Hypothermia was continued for 28 h, followed by rewarming for 12 h. Sedation to a minimum level of Richmond Agitation Sedation Scale (RASS)-4 was mandatory in both intervention groups until 40 h after randomization. Patients received an arterial catheter for hemodynamic monitoring and a CVC was placed with the tip in the superior vena cava. Multimodal neuroprognostication was performed at earliest 96 h after randomization. WLST, due to perceived poor prognosis, was discouraged according to the trial protocol before neuroprognostication unless the patient fulfilled the criteria for complete cerebral infarction.

| Data collection and blood sampling
Data collection included patient demographics, prehospital data, and postcardiac arrest care in accordance with the TTM2 study protocol.
Vasopressor requirements were recorded as extended cardiovascular sequential organ failure assessment (SOFA) score at admission and each day during the initial 7 days for patients remaining in the ICU.
The score ranges from 0 to 8, where 0 is a patient with no vasopressors and a mean arterial pressure above 70 mmHg and 8 is a patient with norepinephrine/epinephrine at >1 μg/kg/min. 17 Arterial blood gases were obtained at randomization and every fourth hour thereafter. Central venous blood gases were obtained simultaneously with arterial blood gases at 4,8,12,16,24,48, and 72 h after randomization. Changes in ventilatory settings were discouraged within 20 min before blood gas sampling to avoid disturbing the ΔpCO 2 steady state. 30 ΔpCO 2 was calculated as the difference between central venous and arterial partial pressure of CO 2 . An elevated ΔpCO 2 was defined as >0.8 kPa and elevated arterial lactate as >4.0 mmoL/L. 18

| Outcomes
The primary outcome was survival status at 96 h after randomization.
This was chosen since the TTM2 protocol discouraged from WLST because of perceived poor neurologic prognosis before this time point. In a sensitivity analysis, we considered a non-neurologic cause of death as the outcome, regardless of time of death. We also analyzed trends and prognostic performance in a subgroup of patients considered to be in shock at admission (Shock was defined according to the TTM2 protocol as a systolic blood pressure of less than 90 mmHg for more than 30 min or end-organ hypoperfusion 29 ).

| Statistical analysis
Continuous variables are presented as mean ± standard deviation (SD) or median with interquartile range (IQR) depending on the distribution of the data. Categorical variables are presented as proportions.
The independent t test and the Mann-Whitney U test were used for the comparison of means and medians, respectively. Paired t tests and Wilcoxon signed-rank tests were used for comparison of means and medians of paired samples. Fisher's exact test was used for comparison of categorical variables. To test for group differences between repeated measures of ΔpCO 2 , lactate, and extended cardiovascular SOFA scores, we used a linear mixed effects model. Logistic regression was used to estimate the odds ratio (OR) between ΔpCO 2 and lactate and 96-h mortality. In the multivariate regression analysis, we adjusted for the following confounders known to be associated with a poor outcome after OHCA: age, initial rhythm, witnessed status, bystander CPR, time to ROSC, and shock at admission (defined according to the TTM2 protocol as a systolic blood pressure of less than 90 mmHg for more than 30 min or end-organ hypoperfusion). 31 In addition, we adjusted for treatment with hypothermia or normothermia. To account for missing data of ΔpCO 2 and lactate, multiple imputation was performed with a linear regression model and five imputed data sets were generated with the Markov Chain Monte Carlo method. Pooled estimates from the imputed data sets were used in the mixed linear effects model. In the multivariate regression analysis, we tested the association between ΔpCO 2 or lactate and 96-h mortality by first including patients with a complete data set and, second, including all patients with pooled estimates from the imputed data sets. The ability of ΔpCO 2 and lactate to separate patients by the outcome of interest was assessed as the area under the receiver operating curve (AUROC). The accuracy of the prognostic performance was classified as follows: an area under the curve (AUC) greater than 0.9 was considered as high, 0.9-0.7 as moderate, and 0.7-0.5 as low. 32 Two-tailed p values <.05 were considered statistically significant. Analyses were conducted using IBM SPSS statistics v27.0 for MAC OS.

| RESULTS
Two hundred seventy one patients were included in the TTM2 trial at the five hospitals participating in this sub-study. One hundred sixtythree patients had at least one measurement of ΔpCO 2 during the initial 24 h after randomization and were included in the present analysis. Measurements at 48 and 72 h were excluded due to a high proportion of missing values (Table S2) and because patients were frequently spontaneously breathing at these time points, which destabilizes the relationship between arterial and venous pCO 2 . All patients had known survival status at 96 h after randomization. Patient characteristics are displayed in Table 1. The characteristics of patients included in the TTM2 trial at the current hospitals, but lacking measurements of ΔpCO 2 are shown in Table S1. The average levels of ΔpCO 2 during the initial 24 h were not different in patients treated with hypothermia versus normothermia. The difference in lactate levels between the groups increased during treatment with hypothermia and was on average higher in patients randomized to treatment with hypothermia ( Figure S1).
The median extended cardiovascular SOFA score at admission, Days 1 and 2, was not different between patients with a high (>0.8 kPa) and normal ΔpCO 2 ( p = .47, Figure S2) Figure S3).
In a sensitivity analysis, we compared trends and prognostic performance of ΔpCO 2 and lactate in patients with and without nonneurologic death (regardless of time of death). Thirty-one patients were classified as having a non-neurologic death, 16 Figure S4).

| DISCUSSION
In the current TTM2 sub-study, our main finding was that average In a previous study by our group, an increase in ΔpCO 2 was associated with a lower risk of ICU mortality and poor neurologic outcome. 28 That study differed from the current study as data were measured at a single time point during the initial 24 h at the discretion of the treating physician. Furthermore, there was a case-mix with patients on extracorporeal membrane oxygenation which possibly affected the relationship between ΔpCO 2 and mortality due to changes in the circulatory physiology with extracorporeal circulatory support. 28 In a study by Riviero et al., there was no association F I G U R E 1 Distribution of ΔpCO 2 (kPa) and lactate (mmol/L) during the initial 24 h after randomization in survivors (blue) and non-survivors (green) at 96 h after admission. In a mixed linear model, the average levels of pΔCO 2 were not significantly different between survivors and non-survivors ( p = .61) but lactate was on average higher in non-survivors during the initial 24 h after randomization ( p < .001).
T A B L E 2 Association between 0.1 kPa ΔpCO 2 increments and 1.0 mmo/L lactate increments at 4 h after randomization with 96-h mortality. between ΔpCO 2 during the initial 72 h and 28-day mortality in postcardiac arrest patients. 27 Further corroboration of our results in the literature is limited due to the fact that post-cardiac arrest patients were not included in the other studies of ΔpCO 2 . [20][21][22][23][24][25]33,34 Although cardiac output was not systematically recorded in our study, we observed a decrease in ΔpCO 2 during the initial 24 h in both survivors and non-survivors. This is in line with previous findings that low cardiac output typically improves within 24 h after ROSC. 4,5,35 Shock at admission was associated with a higher initial ΔpCO 2 but a ΔpCO 2 above 0.8 kPa was not linked to higher vasopressor requirement. Furthermore, ΔpCO 2 was not a good predictor of early death in the subgroup of patients with shock at admission.
The poor prognostic performance of ΔpCO 2 to identify patients with early mortality could pertain to some of its limitations. Changes in ventilation and metabolism may cause rapid fluctuations in ΔpCO 2 independent of tissue perfusion and factors affecting the relationship between pCO 2 and CO 2 content (e.g., pH, hematocrit, and oxygen saturation) may also affect the validity of ΔpCO 2 . 30,36 Furthermore, a low ΔpCO 2 does not rule out ongoing tissue ischemia when cardiac output is normal or high. 19,37,38 This type of shock is more common in the later phase of postresuscitation circulatory failure. 19,35 In contrast to ΔpCO 2 , lactate levels over the initial 24 h were consistently associated with 96-h mortality and lactate had better prognostic performance to identify patients with early mortality. This is possibly explained by the fact that lactate reflects tissue hypoxia both secondary to a low cardiac output state and to a distributive shock with a normal or hyperdynamic cardiac output. 19 There was no difference in 96-h mortality in patients randomized to hypothermia or normothermia and the groups were analyzed together. Hypothermia may theoretically affect ΔpCO 2 due to its effect on metabolism and CO 2 solubility. 41 In the current study, however, levels of ΔpCO 2 did not differ between patients randomized to hypothermia and normothermia. Lactate was higher in patients treated with hypothermia which has been previously demonstrated. 17

| Limitations
Several patients were excluded due to lack of correctly measured ΔpCO 2 . Excluded patients had a higher 96-h mortality and higher levels of lactate at admission, indicating that this group comprised critically ill patients that more frequently died shortly after hospital admission due to circulatory shock. Higher lactate levels at admission could indicate that circulatory shock was more often clinically evident in this group of patients compared to those included in the sub-study.
Although exclusion of these patients is an obvious limitation, we believe that the value of shock biomarkers (i.e., ΔpCO 2 ) is probably smaller in circumstances when circulatory shock is clinically evident.
Our outcome was chosen to capture patients who died secondary to cardiovascular or multiorgan failure. Still, 46% of non-survivors at 96 h were not classified as non-neurologic deaths and would not be expected to be identified by a marker of cardiovascular failure. However, trends and prognostic performance of ΔpCO 2 were largely similar when we used a classification of non-neurologic death as the outcome. Our finding that ΔpCO 2 was associated with the outcome at 4 h should be taken with caution as we did not adjust for multiple testing when we tested the association between our markers and the outcome at multiple time points.
The validity of ΔpCO 2 may be affected by a number of factors.
One important factor is ventilatory changes as both hyper-and hypoventilation has been demonstrated to lead to acute changes in ΔpCO 2 . 30 In the study protocol, sampling was done after 20 min of stable settings in patients with controlled ventilation and we excluded measurements done at 48 and 72 h since many patients were spontaneously breathing at those time points. We lacked physiological parameters such as cardiac output, systemic vascular resistance and measures of microcirculation which would have been valuable in interpreting the results. We used central venous blood rather than mixed venous blood, which may decrease the detection of alterations in venous CO2 from the inferior vena cava.

| CONCLUSION
Our study does not support the use of ΔpCO 2 to identify patients at risk of early mortality after OHCA as there was no consistent difference in ΔpCO 2 levels between survivors and non-survivors in the early postresuscitation phase and initial ΔpCO 2 levels did not identify patients with early mortality. Lactate, however, was consistently higher in non-survivors and at 4 h after randomization lactate identified patients with early mortality with a moderate accuracy.